In [3]:
sns.set(style="darkgrid")
sns.countplot(y= meta_data['class'],orient='v')
plt.show()
import os
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import librosa
import librosa.display
import IPython.display as ipd
from keras.callbacks import Callback
import seaborn as sns
import soundfile as sf
from scipy.io import wavfile as wav
from tqdm import tqdm
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import LabelEncoder
from keras.utils import to_categorical
%matplotlib inline
meta_data = pd.read_csv("UrbanSound8K.csv")
np.shape(meta_data) # (8732, 8)
meta_data.head()
| slice_file_name | fsID | start | end | salience | fold | classID | class | |
|---|---|---|---|---|---|---|---|---|
| 0 | 100032-3-0-0.wav | 100032 | 0.0 | 0.317551 | 1 | 5 | 3 | dog_bark |
| 1 | 100263-2-0-117.wav | 100263 | 58.5 | 62.500000 | 1 | 5 | 2 | children_playing |
| 2 | 100263-2-0-121.wav | 100263 | 60.5 | 64.500000 | 1 | 5 | 2 | children_playing |
| 3 | 100263-2-0-126.wav | 100263 | 63.0 | 67.000000 | 1 | 5 | 2 | children_playing |
| 4 | 100263-2-0-137.wav | 100263 | 68.5 | 72.500000 | 1 | 5 | 2 | children_playing |
sns.set(style="darkgrid")
sns.countplot(y= meta_data['class'],orient='v')
plt.show()
fig = plt.figure(figsize=(15,15))
fig.subplots_adjust(hspace = 0.5, wspace = 0.5)
classes = list(meta_data['class'].unique())
plt.title('Waveplots for different classes of sounds\n',size=20)
plt.axis('off')
for i in range(10):
fig.add_subplot(5,2,i+1)
plt.title(classes[i])
plt.ylabel('Amplitude')
file_name = 'sample_sound/'+classes[i]+".wav"
data,sample_rate = librosa.load(file_name)
librosa.display.waveshow(data,sr=sample_rate)
plt.tight_layout()
# Listening module for a sound class
i = 7
filename = 'sample_sound/'+classes[i]+".wav"
plt.figure(figsize=(10,3))
plt.title(classes[i])
data,sample_rate = librosa.load(filename)
librosa.display.waveshow(data,sr=sample_rate)
ipd.Audio(filename)
# Waveplot of original sound without getting normalized
from scipy.io import wavfile as wav
wave_sample_rate,wave_audio =wav.read(filename)
plt.figure(figsize=(12,5))
plt.plot(wave_audio)
[<matplotlib.lines.Line2D at 0x25315386510>, <matplotlib.lines.Line2D at 0x253153a3450>]
audio_d_list = []
for index, row in meta_data.iterrows():
file_name = os.path.join(os.path.abspath('Audio_dataset/'),
'fold'+ str(row["fold"])+'/',
str(row["slice_file_name"]))
# extracting channels, sampling rate, bit rate and duration
data = sf.SoundFile(file_name)
duration = float(meta_data.iloc[index,3])-float(meta_data.iloc[index,2])
audio_d_list.append([data.channels,
data.samplerate,
data.subtype[4:],
duration])
audio_property = pd.DataFrame(audio_d_list,
columns=['num_channels','sample_rate','bit_depth','duration'])
audio_property.head()
| num_channels | sample_rate | bit_depth | duration | |
|---|---|---|---|---|
| 0 | 2 | 44100 | 16 | 0.317551 |
| 1 | 2 | 44100 | 16 | 4.000000 |
| 2 | 2 | 44100 | 16 | 4.000000 |
| 3 | 2 | 44100 | 16 | 4.000000 |
| 4 | 2 | 44100 | 16 | 4.000000 |
fig = plt.figure(figsize=(20, 8))
fig.subplots_adjust(hspace = 0.5, wspace = 0.5)
classes = list(meta_data['class'].unique())
plt.title('Spectogram for different classes of sounds\n',size=30)
plt.axis('off')
for i in range(10):
fig.add_subplot(2,5,i+1)
plt.title(classes[i])
file_name = 'sample_sound/'+classes[i]+".wav"
data,sample_rate = librosa.load(file_name)
D = librosa.amplitude_to_db(np.abs(librosa.stft(data)), ref=np.max)
librosa.display.specshow(D, x_axis = "time",y_axis='log')
plt.colorbar(format='%+2.0f dB')
plt.tight_layout()
print(audio_property.duration)
0 0.317551
1 4.000000
2 4.000000
3 4.000000
4 4.000000
...
8727 4.000000
8728 2.142545
8729 3.505983
8730 2.532098
8731 2.532099
Name: duration, Length: 8732, dtype: float64
Sound Channel refers to the independent audio signal which is collected or playback when the sound is recording or playback in different spatial position. Therefore, the number of channel is the amount of sound source when the sound is recording or the relevant speaker number when it is playback
print(audio_property.num_channels.value_counts(normalize=True))
2 0.915369 1 0.084631 Name: num_channels, dtype: float64
Bit depth describes the resolution of the sound data that is captured and stored in an audio file. A higher audio bit depth indicates a more detailed sound recording.
print(audio_property.bit_depth.value_counts(normalize=True))
16 0.659414 24 0.315277 T 0.019354 U8 0.004924 DPCM 0.000916 ADPCM 0.000115 Name: bit_depth, dtype: float64
In audio production, a sample rate (or "sampling frequency") defines how many times per second a sound is sampled. Technically speaking, it is the frequency of samples used in a digital recording.
print(audio_property.sample_rate.value_counts(normalize=True))
44100 0.614979 48000 0.286532 96000 0.069858 24000 0.009391 16000 0.005153 22050 0.005039 11025 0.004466 192000 0.001947 8000 0.001374 11024 0.000802 32000 0.000458 Name: sample_rate, dtype: float64
Using Librosa’s load() function:
Using Librosa MFCC function:
def feature_extractor(file_name,scale,padding,max_pad,n):
'''
Function to extract features from the audio file
Does the following things using Librosa library:
- Converts the sampling rate to 22.05 KHz
- Normalize the Bit-depth values of the audio
- Merge the audio channels into a Mono (single) channel
Parameters:
Input:
file_name : string;
Path of the file to be processed
scale : False;
True when 1D features are required calculated
by taking mean along the axis
padding : False;
True when 2D features are required with padding
max_pad : None [int];
Maxium size of a padded image/array.
Required when padding is set to True
n = 40 [int];
Number of MFCCs to return
Output:
mfccs = array of mfccs features.
(1D when scaling = True
2D when padding = True)
'''
try:
audio, sample_rate = librosa.load(file_name, res_type='kaiser_fast')
mfccs = librosa.feature.mfcc(y = audio, sr = sample_rate, n_mfcc = n)
if padding:
pad_width = max_pad - mfccs.shape[1]
mfccs = np.pad(mfccs, pad_width=((0, 0), (0, pad_width)), mode='constant')
if scale:
mfccs = np.mean(mfccs.T,axis=0)
except Exception as e:
print("Error encountered while parsing file: ", file_name)
return None
return mfccs
feat_list = []
for index, row in meta_data.iterrows():
file_name = os.path.join(os.path.abspath('Audio_dataset/'),
'fold'+str(row["fold"])+'/',
str(row["slice_file_name"]))
class_label = row["class"]
data = feature_extractor(file_name,scale=True)
feat_list.append([data, class_label])
features_1D = pd.DataFrame(feat_list, columns=['feature','class_label'])
print('Processed ', len(features_1D), ' files')
print('Output Feature shape ',np.shape(features_1D.iloc[0,0]))
C:\Users\daksh\anaconda3\Lib\site-packages\librosa\core\spectrum.py:256: UserWarning: n_fft=2048 is too large for input signal of length=1323 warnings.warn( C:\Users\daksh\anaconda3\Lib\site-packages\librosa\core\spectrum.py:256: UserWarning: n_fft=2048 is too large for input signal of length=1103 warnings.warn( C:\Users\daksh\anaconda3\Lib\site-packages\librosa\core\spectrum.py:256: UserWarning: n_fft=2048 is too large for input signal of length=1523 warnings.warn(
Processed 8732 files Output Feature shape (40,)
features_1D.to_pickle("features_1D.pkl")
feat_list = []
feat_list_padded = []
for index, row in meta_data.iterrows():
file_name = os.path.join(os.path.abspath('Audio_dataset/'),
'fold'+ str(row["fold"])+'/',
str(row["slice_file_name"]))
class_label = row["class"]
class_id = row["classID"]
data = feature_extractor(file_name)
feat_list.append([data, class_id])
data = feature_extractor(file_name,padding=True,max_pad = 174)
feat_list_padded.append([data, class_label])
features_wo_pad = pd.DataFrame(feat_list, columns=['feature','class_label'])
features_2D = pd.DataFrame(feat_list_padded, columns=['feature','class_label'])
print('Processed ', len(features_2D), ' files')
print('Output Feature Padded shape ',np.shape(features_2D.iloc[0,0]))
C:\Users\daksh\anaconda3\Lib\site-packages\librosa\core\spectrum.py:256: UserWarning: n_fft=2048 is too large for input signal of length=1323 warnings.warn( C:\Users\daksh\anaconda3\Lib\site-packages\librosa\core\spectrum.py:256: UserWarning: n_fft=2048 is too large for input signal of length=1103 warnings.warn( C:\Users\daksh\anaconda3\Lib\site-packages\librosa\core\spectrum.py:256: UserWarning: n_fft=2048 is too large for input signal of length=1523 warnings.warn(
Processed 8732 files Output Feature Padded shape (40, 174)
features_wo_pad.to_pickle("features_wo_pad.pkl")
features_2D.to_pickle("features_2D.pkl")
features_1D = pd.read_pickle("features_1D.pkl")
features_2D = pd.read_pickle("features_2D.pkl")
features_wo_pad = pd.read_pickle("features_wo_pad.pkl")
# choose a sample of sound from each class
sound_class_dic = meta_data.groupby(['class', 'classID']).size()
sound_sample = []
for c in range(10):
for index, row in features_wo_pad.iterrows():
if c == row['class_label']:
sound_sample.append(row)
break
fig = plt.figure(figsize=(20, 8))
fig.subplots_adjust(hspace = 0.5, wspace = 0.5)
classes = list(meta_data['class'].unique())
plt.title('MFCC Spectogram for different classes of sounds\n',size=20)
plt.axis('off')
for i in range(10):
fig.add_subplot(2,5,i+1)
plt.title(classes[i])
plt.ylabel('Amplitude')
file_name = 'sample_sound/'+classes[i]+".wav"
mfcc_ = feature_extractor(file_name)
#mfcc_ = feature_extractor(file_name,padding=True,max_pad=174)
librosa.display.specshow(mfcc_,x_axis='time', y_axis='log')
plt.colorbar(format='%+2.0f dB')
plt.tight_layout()
# Convert features and corresponding classification labels into numpy arrays
X_1D = np.array(features_1D.feature.tolist())
y_1D = np.array(features_1D.class_label.tolist())
# Encode the classification labels
label_encoder = LabelEncoder()
yy_1D = to_categorical(label_encoder.fit_transform(y_1D))
# Convert features and corresponding classification labels into numpy arrays
X_2D = np.array(features_2D.feature.tolist())
y_2D = np.array(features_2D.class_label.tolist())
# Encode the classification labels
yy_2D = to_categorical(label_encoder.transform(y_2D))
x_train_sl_1D, x_test_sl_1D, y_train_sl_1D, y_test_sl_1D = train_test_split(X_1D,y_1D,test_size=0.2,random_state =27)
# x_train_sl_2D, x_test_sl_2D, y_train_sl_2D, y_test_sl_2D = train_test_split(
# X_2D.reshape(X_2D.shape[0],X_2D.shape[1]*X_2D.shape[2]),yy_2D,test_size=0.2,random_state = 42)
x_train_1D, x_test_1D, y_train_1D, y_test_1D = train_test_split(X_1D, yy_1D, test_size=0.2, random_state = 27)
x_train_2D, x_test_2D, y_train_2D, y_test_2D = train_test_split(X_2D, yy_2D, test_size=0.2, random_state = 27)
Following will be used for classification:
from sklearn.ensemble import RandomForestClassifier
from sklearn.datasets import make_classification
clf = RandomForestClassifier(max_depth=4, random_state=0,n_estimators=100)
clf.fit(x_train_sl_1D,y_train_sl_1D)
y_pred = clf.predict(x_test_sl_1D)
train_accu = clf.score(x_train_sl_1D, y_train_sl_1D)
test_accu = clf.score(x_test_sl_1D, y_test_sl_1D)
print('Training Accuracy:', train_accu)
print('Test Accuracy:', test_accu)
Training Accuracy: 0.5712240515390121 Test Accuracy: 0.5334859759587864
import pickle
from sklearn import svm
from sklearn.metrics import roc_curve
from sklearn.metrics import roc_auc_score
from sklearn.metrics import confusion_matrix
from sklearn.utils.multiclass import unique_labels
from sklearn.model_selection import GridSearchCV
import warnings
warnings.filterwarnings('ignore')
#search the hyperparameter space to find the best model
tuned_parameters = [{'kernel': ['rbf'],
'gamma': [1e-3, 1e-4, 1e-5],
'C': [1, 10 ,20,30,40,50]}]
# , {'kernel': ['linear'], 'C': [1, 10, 100, 1000]}]
searchpara = GridSearchCV(svm.SVC(), tuned_parameters)
searchpara.fit(x_train_sl_1D, y_train_sl_1D)
searchpara.best_params_
{'C': 50, 'gamma': 0.0001, 'kernel': 'rbf'}
# construct the model by the best hyperparameters found above
SVM = svm.SVC(C=50.0, gamma=0.001)
# train the model
SVM.fit(x_train_sl_1D,y_train_sl_1D)
y_pred = SVM.predict(x_test_sl_1D)
train_accu = SVM.score(x_train_sl_1D, y_train_sl_1D)
test_accu = SVM.score(x_test_sl_1D, y_test_sl_1D)
print('Training Accuracy: %.4f%%'% (train_accu*100))
print('Test Accuracy: %.4f%%'% (test_accu*100))
Training Accuracy: 99.8712% Test Accuracy: 89.8111%
def plot_confusion_matrix(y_true, y_pred, classes,
normalize=False,
title=None,
cmap=plt.cm.Blues):
"""
This function prints and plots the confusion matrix.
Normalization can be applied by setting `normalize=True`.
"""
if not title:
if normalize:
title = 'Normalized confusion matrix'
else:
title = 'Confusion matrix, without normalization'
# Compute confusion matrix
cm = confusion_matrix(y_true, y_pred)
fig, ax = plt.subplots(figsize = (8,8))
im = ax.imshow(cm, interpolation='nearest', cmap=cmap)
ax.figure.colorbar(im, ax=ax)
ax.set(xticks=np.arange(cm.shape[1]),
yticks=np.arange(cm.shape[0]),
xticklabels=classes, yticklabels=classes,
title=title,
ylabel='True label',
xlabel='Predicted label')
# Rotate the tick labels and set their alignment.
plt.setp(ax.get_xticklabels(), rotation=45, ha="right",
rotation_mode="anchor")
# Loop over data dimensions and create text annotations.
fmt = '.2f' if normalize else 'd'
thresh = cm.max() / 2.
for i in range(cm.shape[0]):
for j in range(cm.shape[1]):
ax.text(j, i, format(cm[i, j], fmt),
ha="center", va="center",
color="white" if cm[i, j] > thresh else "black")
fig.tight_layout()
return ax
class_names = meta_data.groupby(['class', 'classID'], as_index = False).sum()['class']
plot_confusion_matrix(y_test_sl_1D, y_pred, classes=class_names, cmap=plt.cm.Blues)
plt.show()
import numpy as np
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import Convolution2D, Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.layers import Convolution2D, MaxPooling2D
from keras.optimizers import Adam
from keras.utils import to_categorical
from sklearn import metrics
from keras.callbacks import ModelCheckpoint
from datetime import datetime
from sklearn.preprocessing import LabelEncoder
from keras.utils import to_categorical
from sklearn.model_selection import train_test_split
n_labels = yy_1D.shape[1]
in_shape = x_train_1D.shape[1]
n_layers = 2
n_nodes = 100
model_num = 4
# Construct model
model = Sequential()
model.add(Dense(n_nodes, input_shape=(in_shape,)))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(n_nodes))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(n_labels))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', metrics=['accuracy'], optimizer='adam')
model.summary()
# Calculate pre-training accuracy
score = model.evaluate(x_train_1D, y_train_1D, verbose=0)
accuracy = 100*score[1]
print("Pre-training accuracy: %.4f%%" % accuracy)
Model: "sequential_3"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
dense_9 (Dense) (None, 100) 4100
activation_9 (Activation) (None, 100) 0
dropout_6 (Dropout) (None, 100) 0
dense_10 (Dense) (None, 100) 10100
activation_10 (Activation) (None, 100) 0
dropout_7 (Dropout) (None, 100) 0
dense_11 (Dense) (None, 10) 1010
activation_11 (Activation) (None, 10) 0
=================================================================
Total params: 15210 (59.41 KB)
Trainable params: 15210 (59.41 KB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________
Pre-training accuracy: 12.0115%
n_epochs = 100
n_batch = 32
start = datetime.now()
history = model.fit(x_train_1D,
y_train_1D,
batch_size=n_batch,
epochs=n_epochs,
validation_data=(x_test_1D, y_test_1D),
verbose=1)
duration = datetime.now() - start
print("Training finished in time: ", duration)
Epoch 1/100 219/219 [==============================] - 1s 2ms/step - loss: 12.5780 - accuracy: 0.1681 - val_loss: 2.0046 - val_accuracy: 0.3097 Epoch 2/100 219/219 [==============================] - 0s 2ms/step - loss: 2.7194 - accuracy: 0.2301 - val_loss: 1.9956 - val_accuracy: 0.3314 Epoch 3/100 219/219 [==============================] - 0s 2ms/step - loss: 2.1957 - accuracy: 0.2674 - val_loss: 1.8807 - val_accuracy: 0.3749 Epoch 4/100 219/219 [==============================] - 0s 2ms/step - loss: 2.0279 - accuracy: 0.3111 - val_loss: 1.7475 - val_accuracy: 0.4493 Epoch 5/100 219/219 [==============================] - 0s 2ms/step - loss: 1.8812 - accuracy: 0.3569 - val_loss: 1.6417 - val_accuracy: 0.4814 Epoch 6/100 219/219 [==============================] - 0s 2ms/step - loss: 1.7719 - accuracy: 0.3956 - val_loss: 1.5838 - val_accuracy: 0.4991 Epoch 7/100 219/219 [==============================] - 0s 2ms/step - loss: 1.7203 - accuracy: 0.4202 - val_loss: 1.4969 - val_accuracy: 0.5255 Epoch 8/100 219/219 [==============================] - 0s 2ms/step - loss: 1.6524 - accuracy: 0.4445 - val_loss: 1.4528 - val_accuracy: 0.5575 Epoch 9/100 219/219 [==============================] - 0s 2ms/step - loss: 1.6034 - accuracy: 0.4563 - val_loss: 1.3910 - val_accuracy: 0.5793 Epoch 10/100 219/219 [==============================] - 0s 2ms/step - loss: 1.5557 - accuracy: 0.4700 - val_loss: 1.3422 - val_accuracy: 0.5884 Epoch 11/100 219/219 [==============================] - 0s 2ms/step - loss: 1.4870 - accuracy: 0.4958 - val_loss: 1.3057 - val_accuracy: 0.5924 Epoch 12/100 219/219 [==============================] - 0s 2ms/step - loss: 1.4610 - accuracy: 0.5015 - val_loss: 1.2693 - val_accuracy: 0.5850 Epoch 13/100 219/219 [==============================] - 0s 2ms/step - loss: 1.4534 - accuracy: 0.5094 - val_loss: 1.2626 - val_accuracy: 0.5976 Epoch 14/100 219/219 [==============================] - 0s 2ms/step - loss: 1.4118 - accuracy: 0.5180 - val_loss: 1.2119 - val_accuracy: 0.6234 Epoch 15/100 219/219 [==============================] - 0s 2ms/step - loss: 1.3842 - accuracy: 0.5327 - val_loss: 1.2044 - val_accuracy: 0.6234 Epoch 16/100 219/219 [==============================] - 0s 2ms/step - loss: 1.3463 - accuracy: 0.5370 - val_loss: 1.1748 - val_accuracy: 0.6371 Epoch 17/100 219/219 [==============================] - 0s 2ms/step - loss: 1.3389 - accuracy: 0.5403 - val_loss: 1.1498 - val_accuracy: 0.6394 Epoch 18/100 219/219 [==============================] - 0s 2ms/step - loss: 1.2876 - accuracy: 0.5562 - val_loss: 1.1162 - val_accuracy: 0.6560 Epoch 19/100 219/219 [==============================] - 0s 2ms/step - loss: 1.2946 - accuracy: 0.5581 - val_loss: 1.1172 - val_accuracy: 0.6560 Epoch 20/100 219/219 [==============================] - 0s 2ms/step - loss: 1.2570 - accuracy: 0.5744 - val_loss: 1.0794 - val_accuracy: 0.6600 Epoch 21/100 219/219 [==============================] - 0s 2ms/step - loss: 1.2606 - accuracy: 0.5778 - val_loss: 1.0660 - val_accuracy: 0.6726 Epoch 22/100 219/219 [==============================] - 0s 2ms/step - loss: 1.2390 - accuracy: 0.5810 - val_loss: 1.0439 - val_accuracy: 0.6789 Epoch 23/100 219/219 [==============================] - 0s 2ms/step - loss: 1.2198 - accuracy: 0.5810 - val_loss: 1.0328 - val_accuracy: 0.6691 Epoch 24/100 219/219 [==============================] - 0s 2ms/step - loss: 1.2240 - accuracy: 0.5841 - val_loss: 1.0464 - val_accuracy: 0.6651 Epoch 25/100 219/219 [==============================] - 0s 2ms/step - loss: 1.2069 - accuracy: 0.5956 - val_loss: 1.0067 - val_accuracy: 0.6961 Epoch 26/100 219/219 [==============================] - 0s 2ms/step - loss: 1.1954 - accuracy: 0.5891 - val_loss: 1.0037 - val_accuracy: 0.6863 Epoch 27/100 219/219 [==============================] - 0s 2ms/step - loss: 1.1897 - accuracy: 0.6056 - val_loss: 0.9769 - val_accuracy: 0.6966 Epoch 28/100 219/219 [==============================] - 0s 2ms/step - loss: 1.1827 - accuracy: 0.5980 - val_loss: 0.9875 - val_accuracy: 0.6875 Epoch 29/100 219/219 [==============================] - 0s 2ms/step - loss: 1.1441 - accuracy: 0.6057 - val_loss: 0.9743 - val_accuracy: 0.6966 Epoch 30/100 219/219 [==============================] - 0s 2ms/step - loss: 1.1577 - accuracy: 0.6046 - val_loss: 0.9525 - val_accuracy: 0.7109 Epoch 31/100 219/219 [==============================] - 0s 2ms/step - loss: 1.1360 - accuracy: 0.6179 - val_loss: 0.9546 - val_accuracy: 0.6961 Epoch 32/100 219/219 [==============================] - 0s 2ms/step - loss: 1.1392 - accuracy: 0.6119 - val_loss: 0.9558 - val_accuracy: 0.7046 Epoch 33/100 219/219 [==============================] - 0s 2ms/step - loss: 1.1341 - accuracy: 0.6107 - val_loss: 0.9561 - val_accuracy: 0.6829 Epoch 34/100 219/219 [==============================] - 0s 2ms/step - loss: 1.1087 - accuracy: 0.6192 - val_loss: 0.9315 - val_accuracy: 0.7041 Epoch 35/100 219/219 [==============================] - 0s 2ms/step - loss: 1.1519 - accuracy: 0.6082 - val_loss: 0.9393 - val_accuracy: 0.7109 Epoch 36/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0963 - accuracy: 0.6256 - val_loss: 0.9250 - val_accuracy: 0.7132 Epoch 37/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0897 - accuracy: 0.6291 - val_loss: 0.9183 - val_accuracy: 0.7132 Epoch 38/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0940 - accuracy: 0.6203 - val_loss: 0.8933 - val_accuracy: 0.7149 Epoch 39/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0795 - accuracy: 0.6308 - val_loss: 0.9014 - val_accuracy: 0.7115 Epoch 40/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0821 - accuracy: 0.6299 - val_loss: 0.9146 - val_accuracy: 0.7109 Epoch 41/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0714 - accuracy: 0.6329 - val_loss: 0.8945 - val_accuracy: 0.7041 Epoch 42/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0629 - accuracy: 0.6309 - val_loss: 0.8828 - val_accuracy: 0.7167 Epoch 43/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0728 - accuracy: 0.6414 - val_loss: 0.8669 - val_accuracy: 0.7252 Epoch 44/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0434 - accuracy: 0.6417 - val_loss: 0.8562 - val_accuracy: 0.7218 Epoch 45/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0409 - accuracy: 0.6421 - val_loss: 0.8629 - val_accuracy: 0.7247 Epoch 46/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0516 - accuracy: 0.6398 - val_loss: 0.8489 - val_accuracy: 0.7275 Epoch 47/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0489 - accuracy: 0.6451 - val_loss: 0.8673 - val_accuracy: 0.7167 Epoch 48/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0473 - accuracy: 0.6362 - val_loss: 0.8406 - val_accuracy: 0.7373 Epoch 49/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0286 - accuracy: 0.6504 - val_loss: 0.8383 - val_accuracy: 0.7270 Epoch 50/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0204 - accuracy: 0.6515 - val_loss: 0.8405 - val_accuracy: 0.7270 Epoch 51/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0427 - accuracy: 0.6527 - val_loss: 0.8292 - val_accuracy: 0.7373 Epoch 52/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0097 - accuracy: 0.6523 - val_loss: 0.8220 - val_accuracy: 0.7338 Epoch 53/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0374 - accuracy: 0.6498 - val_loss: 0.8176 - val_accuracy: 0.7481 Epoch 54/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9992 - accuracy: 0.6611 - val_loss: 0.8161 - val_accuracy: 0.7470 Epoch 55/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9982 - accuracy: 0.6647 - val_loss: 0.8388 - val_accuracy: 0.7344 Epoch 56/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0177 - accuracy: 0.6535 - val_loss: 0.8282 - val_accuracy: 0.7338 Epoch 57/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0043 - accuracy: 0.6553 - val_loss: 0.8154 - val_accuracy: 0.7396 Epoch 58/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0077 - accuracy: 0.6588 - val_loss: 0.8253 - val_accuracy: 0.7464 Epoch 59/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0020 - accuracy: 0.6588 - val_loss: 0.8136 - val_accuracy: 0.7338 Epoch 60/100 219/219 [==============================] - 0s 2ms/step - loss: 1.0039 - accuracy: 0.6511 - val_loss: 0.8128 - val_accuracy: 0.7384 Epoch 61/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9857 - accuracy: 0.6606 - val_loss: 0.7929 - val_accuracy: 0.7384 Epoch 62/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9760 - accuracy: 0.6686 - val_loss: 0.7910 - val_accuracy: 0.7418 Epoch 63/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9788 - accuracy: 0.6620 - val_loss: 0.7956 - val_accuracy: 0.7533 Epoch 64/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9810 - accuracy: 0.6690 - val_loss: 0.7928 - val_accuracy: 0.7510 Epoch 65/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9629 - accuracy: 0.6674 - val_loss: 0.7810 - val_accuracy: 0.7476 Epoch 66/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9831 - accuracy: 0.6641 - val_loss: 0.8078 - val_accuracy: 0.7333 Epoch 67/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9681 - accuracy: 0.6696 - val_loss: 0.8109 - val_accuracy: 0.7275 Epoch 68/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9710 - accuracy: 0.6653 - val_loss: 0.8009 - val_accuracy: 0.7396 Epoch 69/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9502 - accuracy: 0.6759 - val_loss: 0.7908 - val_accuracy: 0.7413 Epoch 70/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9738 - accuracy: 0.6670 - val_loss: 0.7984 - val_accuracy: 0.7384 Epoch 71/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9809 - accuracy: 0.6624 - val_loss: 0.7876 - val_accuracy: 0.7481 Epoch 72/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9542 - accuracy: 0.6700 - val_loss: 0.7687 - val_accuracy: 0.7436 Epoch 73/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9455 - accuracy: 0.6732 - val_loss: 0.7727 - val_accuracy: 0.7499 Epoch 74/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9734 - accuracy: 0.6649 - val_loss: 0.7809 - val_accuracy: 0.7401 Epoch 75/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9468 - accuracy: 0.6812 - val_loss: 0.7567 - val_accuracy: 0.7613 Epoch 76/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9379 - accuracy: 0.6754 - val_loss: 0.7688 - val_accuracy: 0.7493 Epoch 77/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9397 - accuracy: 0.6816 - val_loss: 0.7466 - val_accuracy: 0.7556 Epoch 78/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9383 - accuracy: 0.6730 - val_loss: 0.7620 - val_accuracy: 0.7533 Epoch 79/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9475 - accuracy: 0.6809 - val_loss: 0.7570 - val_accuracy: 0.7579 Epoch 80/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9647 - accuracy: 0.6639 - val_loss: 0.7554 - val_accuracy: 0.7510 Epoch 81/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9300 - accuracy: 0.6786 - val_loss: 0.7537 - val_accuracy: 0.7510 Epoch 82/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9370 - accuracy: 0.6772 - val_loss: 0.7638 - val_accuracy: 0.7401 Epoch 83/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9314 - accuracy: 0.6817 - val_loss: 0.7516 - val_accuracy: 0.7562 Epoch 84/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9562 - accuracy: 0.6752 - val_loss: 0.7637 - val_accuracy: 0.7539 Epoch 85/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9247 - accuracy: 0.6766 - val_loss: 0.7523 - val_accuracy: 0.7521 Epoch 86/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9329 - accuracy: 0.6826 - val_loss: 0.7452 - val_accuracy: 0.7521 Epoch 87/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9192 - accuracy: 0.6793 - val_loss: 0.7452 - val_accuracy: 0.7436 Epoch 88/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9215 - accuracy: 0.6839 - val_loss: 0.7443 - val_accuracy: 0.7516 Epoch 89/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9046 - accuracy: 0.6833 - val_loss: 0.7283 - val_accuracy: 0.7590 Epoch 90/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9212 - accuracy: 0.6826 - val_loss: 0.7361 - val_accuracy: 0.7619 Epoch 91/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9342 - accuracy: 0.6772 - val_loss: 0.7479 - val_accuracy: 0.7642 Epoch 92/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9222 - accuracy: 0.6832 - val_loss: 0.7349 - val_accuracy: 0.7693 Epoch 93/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9153 - accuracy: 0.6799 - val_loss: 0.7391 - val_accuracy: 0.7527 Epoch 94/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9116 - accuracy: 0.6827 - val_loss: 0.7260 - val_accuracy: 0.7619 Epoch 95/100 219/219 [==============================] - 0s 2ms/step - loss: 0.8983 - accuracy: 0.6929 - val_loss: 0.7193 - val_accuracy: 0.7642 Epoch 96/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9175 - accuracy: 0.6812 - val_loss: 0.7529 - val_accuracy: 0.7459 Epoch 97/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9322 - accuracy: 0.6850 - val_loss: 0.7519 - val_accuracy: 0.7613 Epoch 98/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9071 - accuracy: 0.6869 - val_loss: 0.7319 - val_accuracy: 0.7624 Epoch 99/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9105 - accuracy: 0.6915 - val_loss: 0.7286 - val_accuracy: 0.7647 Epoch 100/100 219/219 [==============================] - 0s 2ms/step - loss: 0.9149 - accuracy: 0.6889 - val_loss: 0.7168 - val_accuracy: 0.7613 Training finished in time: 0:00:40.335810
# Evaluating the model on the training and testing set
acc_train = model.evaluate(x_train_1D, y_train_1D, verbose=0)
print("Training Accuracy: ", acc_train[1])
acc_test = model.evaluate(x_test_1D, y_test_1D, verbose=0)
print("Testing Accuracy: ", acc_test[1])
Training Accuracy: 0.8210451006889343 Testing Accuracy: 0.7613050937652588
def plot_accuracy(history,n_epochs,n_batch):
training_accuracy = history.history['accuracy']
testing_accuracy = history.history['val_accuracy'] # Assuming you used 'val_accuracy' for testing accuracy
# Create a list of epoch numbers
epochs = range(1, n_epochs + 1)
# Plot both training and testing accuracy
plt.plot(epochs, training_accuracy, 'b', label='Training Accuracy')
plt.plot(epochs, testing_accuracy, 'r', label='Testing Accuracy')
plt.title('Accuracy vs Epochs')
plt.xlabel('Epochs')
plt.ylabel('Accuracy')
plt.legend()
plt.show()
def plot_losses(history,n_epochs):
train_losses = history.history['loss']
test_losses = history.history['val_loss'] # Assuming you used 'val_loss' for validation loss
# Create a list of epoch numbers
epochs = range(1, len(train_losses) + 1)
# Plot training and validation losses
plt.plot(epochs, train_losses, 'b', label='Training Loss')
plt.plot(epochs, test_losses, 'r', label='Validation Loss')
plt.title('Training and Testing Loss vs Epochs')
plt.xlabel('Epochs')
plt.ylabel('Loss')
plt.legend()
plt.show()
plot_accuracy(history,n_epochs,32)
model_3 = Sequential()
model_3.add(Dense(n_nodes, input_shape=(in_shape,)))
model_3.add(Activation('relu'))
model_3.add(Dropout(0.5))
model_3.add(Dense(n_nodes))
model_3.add(Activation('relu'))
model_3.add(Dropout(0.5))
model_3.add(Dense(n_labels))
model_3.add(Activation('softmax'))
model_3.compile(loss='categorical_crossentropy', metrics=['accuracy'], optimizer='adam')
n_epochs_2=200
n_batch_2=16
start = datetime.now()
history = model_3.fit(x_train_1D,
y_train_1D,
batch_size=n_batch_2,
epochs=n_epochs_2,
validation_data=(x_test_1D, y_test_1D),
verbose=1)
duration = datetime.now() - start
print("Training finished in time: ", duration)
Epoch 1/200 437/437 [==============================] - 1s 2ms/step - loss: 9.9619 - accuracy: 0.1482 - val_loss: 2.1924 - val_accuracy: 0.1540 Epoch 2/200 437/437 [==============================] - 1s 1ms/step - loss: 2.4366 - accuracy: 0.1585 - val_loss: 2.1705 - val_accuracy: 0.1998 Epoch 3/200 437/437 [==============================] - 1s 1ms/step - loss: 2.2086 - accuracy: 0.2113 - val_loss: 2.0403 - val_accuracy: 0.2685 Epoch 4/200 437/437 [==============================] - 1s 2ms/step - loss: 2.0906 - accuracy: 0.2517 - val_loss: 1.9333 - val_accuracy: 0.3331 Epoch 5/200 437/437 [==============================] - 1s 1ms/step - loss: 1.9878 - accuracy: 0.2775 - val_loss: 1.8224 - val_accuracy: 0.3578 Epoch 6/200 437/437 [==============================] - 1s 1ms/step - loss: 1.8943 - accuracy: 0.3152 - val_loss: 1.7304 - val_accuracy: 0.4001 Epoch 7/200 437/437 [==============================] - 1s 1ms/step - loss: 1.8242 - accuracy: 0.3396 - val_loss: 1.6243 - val_accuracy: 0.4322 Epoch 8/200 437/437 [==============================] - 1s 1ms/step - loss: 1.7620 - accuracy: 0.3601 - val_loss: 1.5546 - val_accuracy: 0.4791 Epoch 9/200 437/437 [==============================] - 1s 1ms/step - loss: 1.6750 - accuracy: 0.4059 - val_loss: 1.4558 - val_accuracy: 0.5054 Epoch 10/200 437/437 [==============================] - 1s 2ms/step - loss: 1.6208 - accuracy: 0.4319 - val_loss: 1.4183 - val_accuracy: 0.5358 Epoch 11/200 437/437 [==============================] - 1s 1ms/step - loss: 1.5447 - accuracy: 0.4583 - val_loss: 1.3594 - val_accuracy: 0.5718 Epoch 12/200 437/437 [==============================] - 1s 1ms/step - loss: 1.5139 - accuracy: 0.4747 - val_loss: 1.2894 - val_accuracy: 0.5867 Epoch 13/200 437/437 [==============================] - 1s 1ms/step - loss: 1.4425 - accuracy: 0.4995 - val_loss: 1.3046 - val_accuracy: 0.5770 Epoch 14/200 437/437 [==============================] - 1s 1ms/step - loss: 1.4209 - accuracy: 0.5064 - val_loss: 1.2083 - val_accuracy: 0.6050 Epoch 15/200 437/437 [==============================] - 1s 1ms/step - loss: 1.3855 - accuracy: 0.5284 - val_loss: 1.1875 - val_accuracy: 0.6050 Epoch 16/200 437/437 [==============================] - 1s 1ms/step - loss: 1.3570 - accuracy: 0.5261 - val_loss: 1.1722 - val_accuracy: 0.6113 Epoch 17/200 437/437 [==============================] - 1s 1ms/step - loss: 1.3275 - accuracy: 0.5435 - val_loss: 1.1406 - val_accuracy: 0.6394 Epoch 18/200 437/437 [==============================] - 1s 1ms/step - loss: 1.2885 - accuracy: 0.5493 - val_loss: 1.0723 - val_accuracy: 0.6554 Epoch 19/200 437/437 [==============================] - 1s 1ms/step - loss: 1.2837 - accuracy: 0.5573 - val_loss: 1.0897 - val_accuracy: 0.6388 Epoch 20/200 437/437 [==============================] - 1s 2ms/step - loss: 1.2587 - accuracy: 0.5724 - val_loss: 1.0526 - val_accuracy: 0.6640 Epoch 21/200 437/437 [==============================] - 1s 1ms/step - loss: 1.2478 - accuracy: 0.5724 - val_loss: 1.0464 - val_accuracy: 0.6646 Epoch 22/200 437/437 [==============================] - 1s 2ms/step - loss: 1.2317 - accuracy: 0.5780 - val_loss: 1.0351 - val_accuracy: 0.6600 Epoch 23/200 437/437 [==============================] - 1s 2ms/step - loss: 1.2059 - accuracy: 0.5863 - val_loss: 0.9953 - val_accuracy: 0.6835 Epoch 24/200 437/437 [==============================] - 1s 2ms/step - loss: 1.1812 - accuracy: 0.5908 - val_loss: 0.9994 - val_accuracy: 0.6800 Epoch 25/200 437/437 [==============================] - 1s 2ms/step - loss: 1.1484 - accuracy: 0.6063 - val_loss: 0.9770 - val_accuracy: 0.6789 Epoch 26/200 437/437 [==============================] - 1s 1ms/step - loss: 1.1527 - accuracy: 0.6053 - val_loss: 0.9488 - val_accuracy: 0.7012 Epoch 27/200 437/437 [==============================] - 1s 1ms/step - loss: 1.1473 - accuracy: 0.6060 - val_loss: 0.9336 - val_accuracy: 0.7046 Epoch 28/200 437/437 [==============================] - 1s 1ms/step - loss: 1.1467 - accuracy: 0.6126 - val_loss: 0.9282 - val_accuracy: 0.6961 Epoch 29/200 437/437 [==============================] - 1s 2ms/step - loss: 1.1452 - accuracy: 0.6175 - val_loss: 0.9261 - val_accuracy: 0.6949 Epoch 30/200 437/437 [==============================] - 1s 2ms/step - loss: 1.1235 - accuracy: 0.6122 - val_loss: 0.9123 - val_accuracy: 0.7001 Epoch 31/200 437/437 [==============================] - 1s 2ms/step - loss: 1.1045 - accuracy: 0.6202 - val_loss: 0.8977 - val_accuracy: 0.7132 Epoch 32/200 437/437 [==============================] - 1s 2ms/step - loss: 1.1008 - accuracy: 0.6245 - val_loss: 0.9018 - val_accuracy: 0.7029 Epoch 33/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0949 - accuracy: 0.6249 - val_loss: 0.8856 - val_accuracy: 0.7144 Epoch 34/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0980 - accuracy: 0.6311 - val_loss: 0.8738 - val_accuracy: 0.7149 Epoch 35/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0660 - accuracy: 0.6321 - val_loss: 0.8429 - val_accuracy: 0.7235 Epoch 36/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0930 - accuracy: 0.6328 - val_loss: 0.8514 - val_accuracy: 0.7264 Epoch 37/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0715 - accuracy: 0.6321 - val_loss: 0.8491 - val_accuracy: 0.7201 Epoch 38/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0426 - accuracy: 0.6427 - val_loss: 0.8751 - val_accuracy: 0.7195 Epoch 39/200 437/437 [==============================] - 1s 2ms/step - loss: 1.0667 - accuracy: 0.6349 - val_loss: 0.8372 - val_accuracy: 0.7207 Epoch 40/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0654 - accuracy: 0.6366 - val_loss: 0.8339 - val_accuracy: 0.7275 Epoch 41/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0338 - accuracy: 0.6523 - val_loss: 0.8306 - val_accuracy: 0.7367 Epoch 42/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0526 - accuracy: 0.6424 - val_loss: 0.8430 - val_accuracy: 0.7264 Epoch 43/200 437/437 [==============================] - 1s 2ms/step - loss: 1.0404 - accuracy: 0.6467 - val_loss: 0.8435 - val_accuracy: 0.7350 Epoch 44/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0521 - accuracy: 0.6407 - val_loss: 0.8198 - val_accuracy: 0.7401 Epoch 45/200 437/437 [==============================] - 1s 2ms/step - loss: 1.0201 - accuracy: 0.6518 - val_loss: 0.8396 - val_accuracy: 0.7281 Epoch 46/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0442 - accuracy: 0.6485 - val_loss: 0.8298 - val_accuracy: 0.7350 Epoch 47/200 437/437 [==============================] - 1s 2ms/step - loss: 1.0224 - accuracy: 0.6533 - val_loss: 0.8242 - val_accuracy: 0.7224 Epoch 48/200 437/437 [==============================] - 1s 2ms/step - loss: 1.0336 - accuracy: 0.6624 - val_loss: 0.8165 - val_accuracy: 0.7310 Epoch 49/200 437/437 [==============================] - 1s 2ms/step - loss: 1.0145 - accuracy: 0.6620 - val_loss: 0.8093 - val_accuracy: 0.7315 Epoch 50/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0061 - accuracy: 0.6555 - val_loss: 0.8122 - val_accuracy: 0.7333 Epoch 51/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0078 - accuracy: 0.6623 - val_loss: 0.8009 - val_accuracy: 0.7464 Epoch 52/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0033 - accuracy: 0.6571 - val_loss: 0.8028 - val_accuracy: 0.7441 Epoch 53/200 437/437 [==============================] - 1s 1ms/step - loss: 1.0028 - accuracy: 0.6591 - val_loss: 0.8304 - val_accuracy: 0.7275 Epoch 54/200 437/437 [==============================] - 1s 2ms/step - loss: 1.0017 - accuracy: 0.6557 - val_loss: 0.7900 - val_accuracy: 0.7350 Epoch 55/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9987 - accuracy: 0.6584 - val_loss: 0.7971 - val_accuracy: 0.7424 Epoch 56/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9953 - accuracy: 0.6628 - val_loss: 0.7932 - val_accuracy: 0.7338 Epoch 57/200 437/437 [==============================] - 1s 2ms/step - loss: 1.0013 - accuracy: 0.6617 - val_loss: 0.7953 - val_accuracy: 0.7441 Epoch 58/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9672 - accuracy: 0.6716 - val_loss: 0.7649 - val_accuracy: 0.7527 Epoch 59/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9794 - accuracy: 0.6614 - val_loss: 0.7906 - val_accuracy: 0.7413 Epoch 60/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9981 - accuracy: 0.6639 - val_loss: 0.7792 - val_accuracy: 0.7436 Epoch 61/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9876 - accuracy: 0.6670 - val_loss: 0.7571 - val_accuracy: 0.7401 Epoch 62/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9770 - accuracy: 0.6727 - val_loss: 0.7845 - val_accuracy: 0.7436 Epoch 63/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9852 - accuracy: 0.6681 - val_loss: 0.7910 - val_accuracy: 0.7493 Epoch 64/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9589 - accuracy: 0.6770 - val_loss: 0.7799 - val_accuracy: 0.7544 Epoch 65/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9777 - accuracy: 0.6680 - val_loss: 0.7779 - val_accuracy: 0.7350 Epoch 66/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9812 - accuracy: 0.6673 - val_loss: 0.7628 - val_accuracy: 0.7533 Epoch 67/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9752 - accuracy: 0.6730 - val_loss: 0.7823 - val_accuracy: 0.7470 Epoch 68/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9804 - accuracy: 0.6699 - val_loss: 0.7445 - val_accuracy: 0.7619 Epoch 69/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9674 - accuracy: 0.6684 - val_loss: 0.7546 - val_accuracy: 0.7670 Epoch 70/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9723 - accuracy: 0.6727 - val_loss: 0.7679 - val_accuracy: 0.7550 Epoch 71/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9706 - accuracy: 0.6681 - val_loss: 0.7638 - val_accuracy: 0.7682 Epoch 72/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9564 - accuracy: 0.6737 - val_loss: 0.7742 - val_accuracy: 0.7499 Epoch 73/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9501 - accuracy: 0.6785 - val_loss: 0.7622 - val_accuracy: 0.7630 Epoch 74/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9583 - accuracy: 0.6773 - val_loss: 0.7530 - val_accuracy: 0.7544 Epoch 75/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9739 - accuracy: 0.6709 - val_loss: 0.7670 - val_accuracy: 0.7550 Epoch 76/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9756 - accuracy: 0.6750 - val_loss: 0.7613 - val_accuracy: 0.7516 Epoch 77/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9464 - accuracy: 0.6773 - val_loss: 0.7707 - val_accuracy: 0.7556 Epoch 78/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9421 - accuracy: 0.6769 - val_loss: 0.7329 - val_accuracy: 0.7636 Epoch 79/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9630 - accuracy: 0.6757 - val_loss: 0.7445 - val_accuracy: 0.7596 Epoch 80/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9511 - accuracy: 0.6742 - val_loss: 0.7286 - val_accuracy: 0.7773 Epoch 81/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9386 - accuracy: 0.6832 - val_loss: 0.7682 - val_accuracy: 0.7436 Epoch 82/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9212 - accuracy: 0.6838 - val_loss: 0.7404 - val_accuracy: 0.7630 Epoch 83/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9793 - accuracy: 0.6723 - val_loss: 0.7497 - val_accuracy: 0.7567 Epoch 84/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9286 - accuracy: 0.6829 - val_loss: 0.7444 - val_accuracy: 0.7539 Epoch 85/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9402 - accuracy: 0.6819 - val_loss: 0.7347 - val_accuracy: 0.7584 Epoch 86/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9340 - accuracy: 0.6879 - val_loss: 0.7473 - val_accuracy: 0.7441 Epoch 87/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9458 - accuracy: 0.6797 - val_loss: 0.7381 - val_accuracy: 0.7642 Epoch 88/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9221 - accuracy: 0.6853 - val_loss: 0.7373 - val_accuracy: 0.7607 Epoch 89/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9140 - accuracy: 0.6859 - val_loss: 0.7482 - val_accuracy: 0.7590 Epoch 90/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9291 - accuracy: 0.6873 - val_loss: 0.7345 - val_accuracy: 0.7682 Epoch 91/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9354 - accuracy: 0.6830 - val_loss: 0.7169 - val_accuracy: 0.7716 Epoch 92/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9271 - accuracy: 0.6797 - val_loss: 0.7062 - val_accuracy: 0.7722 Epoch 93/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9313 - accuracy: 0.6825 - val_loss: 0.7360 - val_accuracy: 0.7584 Epoch 94/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9332 - accuracy: 0.6826 - val_loss: 0.7314 - val_accuracy: 0.7693 Epoch 95/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9296 - accuracy: 0.6812 - val_loss: 0.7228 - val_accuracy: 0.7699 Epoch 96/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9359 - accuracy: 0.6817 - val_loss: 0.7366 - val_accuracy: 0.7636 Epoch 97/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9423 - accuracy: 0.6777 - val_loss: 0.7305 - val_accuracy: 0.7590 Epoch 98/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9150 - accuracy: 0.6901 - val_loss: 0.7372 - val_accuracy: 0.7579 Epoch 99/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9225 - accuracy: 0.6892 - val_loss: 0.7422 - val_accuracy: 0.7579 Epoch 100/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9311 - accuracy: 0.6883 - val_loss: 0.7354 - val_accuracy: 0.7556 Epoch 101/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9406 - accuracy: 0.6882 - val_loss: 0.7453 - val_accuracy: 0.7596 Epoch 102/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9218 - accuracy: 0.6853 - val_loss: 0.7224 - val_accuracy: 0.7693 Epoch 103/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9037 - accuracy: 0.6905 - val_loss: 0.7042 - val_accuracy: 0.7768 Epoch 104/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9244 - accuracy: 0.6931 - val_loss: 0.7055 - val_accuracy: 0.7705 Epoch 105/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9261 - accuracy: 0.6911 - val_loss: 0.7096 - val_accuracy: 0.7710 Epoch 106/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9340 - accuracy: 0.6852 - val_loss: 0.7301 - val_accuracy: 0.7630 Epoch 107/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9171 - accuracy: 0.6923 - val_loss: 0.7028 - val_accuracy: 0.7728 Epoch 108/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8986 - accuracy: 0.6989 - val_loss: 0.7114 - val_accuracy: 0.7693 Epoch 109/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9233 - accuracy: 0.6893 - val_loss: 0.7068 - val_accuracy: 0.7768 Epoch 110/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9063 - accuracy: 0.6976 - val_loss: 0.6842 - val_accuracy: 0.7939 Epoch 111/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9077 - accuracy: 0.6951 - val_loss: 0.7232 - val_accuracy: 0.7607 Epoch 112/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9103 - accuracy: 0.6921 - val_loss: 0.6987 - val_accuracy: 0.7624 Epoch 113/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9167 - accuracy: 0.6971 - val_loss: 0.7337 - val_accuracy: 0.7544 Epoch 114/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9193 - accuracy: 0.6966 - val_loss: 0.7140 - val_accuracy: 0.7613 Epoch 115/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9166 - accuracy: 0.6903 - val_loss: 0.7119 - val_accuracy: 0.7596 Epoch 116/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9201 - accuracy: 0.6875 - val_loss: 0.7415 - val_accuracy: 0.7573 Epoch 117/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9224 - accuracy: 0.6999 - val_loss: 0.7100 - val_accuracy: 0.7733 Epoch 118/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8951 - accuracy: 0.7034 - val_loss: 0.7055 - val_accuracy: 0.7825 Epoch 119/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8864 - accuracy: 0.7001 - val_loss: 0.7120 - val_accuracy: 0.7745 Epoch 120/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8993 - accuracy: 0.6911 - val_loss: 0.7055 - val_accuracy: 0.7779 Epoch 121/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9095 - accuracy: 0.6921 - val_loss: 0.7111 - val_accuracy: 0.7790 Epoch 122/200 437/437 [==============================] - 1s 2ms/step - loss: 0.9022 - accuracy: 0.7004 - val_loss: 0.6778 - val_accuracy: 0.7848 Epoch 123/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9098 - accuracy: 0.6938 - val_loss: 0.7226 - val_accuracy: 0.7602 Epoch 124/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9153 - accuracy: 0.6974 - val_loss: 0.7266 - val_accuracy: 0.7653 Epoch 125/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9182 - accuracy: 0.6931 - val_loss: 0.7096 - val_accuracy: 0.7762 Epoch 126/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8994 - accuracy: 0.6984 - val_loss: 0.7038 - val_accuracy: 0.7665 Epoch 127/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9033 - accuracy: 0.6929 - val_loss: 0.7100 - val_accuracy: 0.7762 Epoch 128/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9034 - accuracy: 0.6946 - val_loss: 0.7236 - val_accuracy: 0.7533 Epoch 129/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8646 - accuracy: 0.7051 - val_loss: 0.6910 - val_accuracy: 0.7636 Epoch 130/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9057 - accuracy: 0.6931 - val_loss: 0.6915 - val_accuracy: 0.7739 Epoch 131/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8847 - accuracy: 0.6974 - val_loss: 0.7022 - val_accuracy: 0.7790 Epoch 132/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9010 - accuracy: 0.6905 - val_loss: 0.7161 - val_accuracy: 0.7647 Epoch 133/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8961 - accuracy: 0.6982 - val_loss: 0.7090 - val_accuracy: 0.7642 Epoch 134/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8994 - accuracy: 0.6985 - val_loss: 0.6837 - val_accuracy: 0.7773 Epoch 135/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8830 - accuracy: 0.7057 - val_loss: 0.7301 - val_accuracy: 0.7619 Epoch 136/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9032 - accuracy: 0.7006 - val_loss: 0.7083 - val_accuracy: 0.7682 Epoch 137/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8739 - accuracy: 0.7041 - val_loss: 0.7129 - val_accuracy: 0.7533 Epoch 138/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8650 - accuracy: 0.7059 - val_loss: 0.7208 - val_accuracy: 0.7705 Epoch 139/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8976 - accuracy: 0.6925 - val_loss: 0.7124 - val_accuracy: 0.7607 Epoch 140/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8849 - accuracy: 0.7008 - val_loss: 0.6755 - val_accuracy: 0.7705 Epoch 141/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9108 - accuracy: 0.6933 - val_loss: 0.7113 - val_accuracy: 0.7705 Epoch 142/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9011 - accuracy: 0.7004 - val_loss: 0.7048 - val_accuracy: 0.7722 Epoch 143/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8940 - accuracy: 0.7015 - val_loss: 0.6901 - val_accuracy: 0.7745 Epoch 144/200 437/437 [==============================] - 1s 2ms/step - loss: 0.8990 - accuracy: 0.6949 - val_loss: 0.6810 - val_accuracy: 0.7705 Epoch 145/200 437/437 [==============================] - 1s 2ms/step - loss: 0.8984 - accuracy: 0.6935 - val_loss: 0.6917 - val_accuracy: 0.7750 Epoch 146/200 437/437 [==============================] - 1s 2ms/step - loss: 0.8720 - accuracy: 0.7049 - val_loss: 0.6949 - val_accuracy: 0.7756 Epoch 147/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8869 - accuracy: 0.7011 - val_loss: 0.6902 - val_accuracy: 0.7728 Epoch 148/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8845 - accuracy: 0.7002 - val_loss: 0.6975 - val_accuracy: 0.7819 Epoch 149/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8795 - accuracy: 0.7075 - val_loss: 0.6903 - val_accuracy: 0.7796 Epoch 150/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8883 - accuracy: 0.7065 - val_loss: 0.6962 - val_accuracy: 0.7779 Epoch 151/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8949 - accuracy: 0.6955 - val_loss: 0.6894 - val_accuracy: 0.7750 Epoch 152/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8783 - accuracy: 0.7032 - val_loss: 0.6984 - val_accuracy: 0.7762 Epoch 153/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9086 - accuracy: 0.6935 - val_loss: 0.6934 - val_accuracy: 0.7728 Epoch 154/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8913 - accuracy: 0.6955 - val_loss: 0.6892 - val_accuracy: 0.7722 Epoch 155/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9185 - accuracy: 0.6974 - val_loss: 0.7108 - val_accuracy: 0.7728 Epoch 156/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8770 - accuracy: 0.7054 - val_loss: 0.7096 - val_accuracy: 0.7653 Epoch 157/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8814 - accuracy: 0.7001 - val_loss: 0.7082 - val_accuracy: 0.7647 Epoch 158/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8942 - accuracy: 0.6981 - val_loss: 0.7145 - val_accuracy: 0.7624 Epoch 159/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8826 - accuracy: 0.7068 - val_loss: 0.7130 - val_accuracy: 0.7630 Epoch 160/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8615 - accuracy: 0.7112 - val_loss: 0.6788 - val_accuracy: 0.7905 Epoch 161/200 437/437 [==============================] - 1s 1ms/step - loss: 0.9068 - accuracy: 0.6996 - val_loss: 0.7017 - val_accuracy: 0.7653 Epoch 162/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8746 - accuracy: 0.7014 - val_loss: 0.6693 - val_accuracy: 0.7899 Epoch 163/200 437/437 [==============================] - 1s 2ms/step - loss: 0.8788 - accuracy: 0.7107 - val_loss: 0.7056 - val_accuracy: 0.7773 Epoch 164/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8868 - accuracy: 0.7037 - val_loss: 0.6873 - val_accuracy: 0.7756 Epoch 165/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8854 - accuracy: 0.7048 - val_loss: 0.6780 - val_accuracy: 0.7779 Epoch 166/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8788 - accuracy: 0.7032 - val_loss: 0.7029 - val_accuracy: 0.7699 Epoch 167/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8840 - accuracy: 0.7016 - val_loss: 0.6853 - val_accuracy: 0.7756 Epoch 168/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8667 - accuracy: 0.7078 - val_loss: 0.6947 - val_accuracy: 0.7728 Epoch 169/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8815 - accuracy: 0.7032 - val_loss: 0.6883 - val_accuracy: 0.7779 Epoch 170/200 437/437 [==============================] - 1s 2ms/step - loss: 0.8944 - accuracy: 0.6979 - val_loss: 0.6810 - val_accuracy: 0.7790 Epoch 171/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8565 - accuracy: 0.7155 - val_loss: 0.6824 - val_accuracy: 0.7722 Epoch 172/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8663 - accuracy: 0.7068 - val_loss: 0.6857 - val_accuracy: 0.7745 Epoch 173/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8739 - accuracy: 0.7105 - val_loss: 0.7058 - val_accuracy: 0.7682 Epoch 174/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8616 - accuracy: 0.7127 - val_loss: 0.6788 - val_accuracy: 0.7745 Epoch 175/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8579 - accuracy: 0.7088 - val_loss: 0.7036 - val_accuracy: 0.7665 Epoch 176/200 437/437 [==============================] - 1s 2ms/step - loss: 0.8759 - accuracy: 0.7051 - val_loss: 0.6807 - val_accuracy: 0.7768 Epoch 177/200 437/437 [==============================] - 1s 2ms/step - loss: 0.8721 - accuracy: 0.7054 - val_loss: 0.6797 - val_accuracy: 0.7745 Epoch 178/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8514 - accuracy: 0.7184 - val_loss: 0.6863 - val_accuracy: 0.7699 Epoch 179/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8601 - accuracy: 0.7102 - val_loss: 0.6857 - val_accuracy: 0.7773 Epoch 180/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8809 - accuracy: 0.7022 - val_loss: 0.6794 - val_accuracy: 0.7819 Epoch 181/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8728 - accuracy: 0.7031 - val_loss: 0.7144 - val_accuracy: 0.7539 Epoch 182/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8796 - accuracy: 0.7031 - val_loss: 0.6798 - val_accuracy: 0.7819 Epoch 183/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8783 - accuracy: 0.7104 - val_loss: 0.6777 - val_accuracy: 0.7768 Epoch 184/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8567 - accuracy: 0.7028 - val_loss: 0.6543 - val_accuracy: 0.7836 Epoch 185/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8721 - accuracy: 0.7132 - val_loss: 0.6719 - val_accuracy: 0.7716 Epoch 186/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8785 - accuracy: 0.7092 - val_loss: 0.6882 - val_accuracy: 0.7653 Epoch 187/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8702 - accuracy: 0.7032 - val_loss: 0.6849 - val_accuracy: 0.7733 Epoch 188/200 437/437 [==============================] - 1s 2ms/step - loss: 0.8508 - accuracy: 0.7134 - val_loss: 0.6769 - val_accuracy: 0.7779 Epoch 189/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8765 - accuracy: 0.7044 - val_loss: 0.6891 - val_accuracy: 0.7728 Epoch 190/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8586 - accuracy: 0.7081 - val_loss: 0.6690 - val_accuracy: 0.7779 Epoch 191/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8567 - accuracy: 0.7178 - val_loss: 0.6808 - val_accuracy: 0.7739 Epoch 192/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8826 - accuracy: 0.7005 - val_loss: 0.6881 - val_accuracy: 0.7808 Epoch 193/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8554 - accuracy: 0.7112 - val_loss: 0.6963 - val_accuracy: 0.7687 Epoch 194/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8674 - accuracy: 0.7038 - val_loss: 0.6826 - val_accuracy: 0.7682 Epoch 195/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8690 - accuracy: 0.7051 - val_loss: 0.7196 - val_accuracy: 0.7590 Epoch 196/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8856 - accuracy: 0.7055 - val_loss: 0.6802 - val_accuracy: 0.7779 Epoch 197/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8835 - accuracy: 0.7044 - val_loss: 0.6797 - val_accuracy: 0.7790 Epoch 198/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8591 - accuracy: 0.7075 - val_loss: 0.6875 - val_accuracy: 0.7693 Epoch 199/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8766 - accuracy: 0.7094 - val_loss: 0.6744 - val_accuracy: 0.7739 Epoch 200/200 437/437 [==============================] - 1s 1ms/step - loss: 0.8797 - accuracy: 0.7042 - val_loss: 0.6686 - val_accuracy: 0.7785 Training finished in time: 0:02:09.654524
plot_accuracy(history,n_epochs_2,n_batch_2)
plot_losses(history,n_epochs_2)
import numpy as np
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation, Flatten
from keras.layers import Convolution2D, Conv2D, MaxPooling2D, GlobalAveragePooling2D
from keras.layers import Convolution2D, MaxPooling2D
from keras.optimizers import Adam
from keras.utils import to_categorical
from sklearn import metrics
from keras.callbacks import ModelCheckpoint
from datetime import datetime
from sklearn.preprocessing import LabelEncoder
from keras.utils import to_categorical
from sklearn.model_selection import train_test_split
from keras.models import model_from_json
num_rows = 40
num_columns = 174
num_channels = 1
x_train_2D = x_train_2D.reshape(x_train_2D.shape[0], num_rows, num_columns, num_channels)
x_test_2D = x_test_2D.reshape(x_test_2D.shape[0], num_rows, num_columns, num_channels)
num_labels = yy_2D.shape[1]
filter_size = 2
# Model Configuration
cnn_model_num = 5
n_labels = yy_1D.shape[1]
n_conv_layers = 4
n_mlp_layers = 2
model = Sequential()
model.add(Conv2D(filters=16, kernel_size=2, input_shape=(num_rows, num_columns, num_channels), activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Dropout(0.2))
model.add(Conv2D(filters=32, kernel_size=2, activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Dropout(0.2))
model.add(Conv2D(filters=64, kernel_size=2, activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Dropout(0.2))
model.add(Conv2D(filters=128, kernel_size=2, activation='relu'))
model.add(MaxPooling2D(pool_size=2))
model.add(Dropout(0.2))
#model.add(GlobalAveragePooling2D())
model.add(Flatten())
model.add(Dense(256))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(256))
model.add(Activation('relu'))
model.add(Dropout(0.5))
model.add(Dense(n_labels))
model.add(Activation('softmax'))
model.compile(loss='categorical_crossentropy', metrics=['accuracy'], optimizer='adam')
model.summary()
# Calculate pre-training accuracy
score = model.evaluate(x_test_2D, y_test_2D, verbose=1)
accuracy = 100*score[1]
print("Pre-training accuracy: %.4f%%" % accuracy)
Model: "sequential_6"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv2d (Conv2D) (None, 39, 173, 16) 80
max_pooling2d (MaxPooling2 (None, 19, 86, 16) 0
D)
dropout_12 (Dropout) (None, 19, 86, 16) 0
conv2d_1 (Conv2D) (None, 18, 85, 32) 2080
max_pooling2d_1 (MaxPoolin (None, 9, 42, 32) 0
g2D)
dropout_13 (Dropout) (None, 9, 42, 32) 0
conv2d_2 (Conv2D) (None, 8, 41, 64) 8256
max_pooling2d_2 (MaxPoolin (None, 4, 20, 64) 0
g2D)
dropout_14 (Dropout) (None, 4, 20, 64) 0
conv2d_3 (Conv2D) (None, 3, 19, 128) 32896
max_pooling2d_3 (MaxPoolin (None, 1, 9, 128) 0
g2D)
dropout_15 (Dropout) (None, 1, 9, 128) 0
flatten (Flatten) (None, 1152) 0
dense_18 (Dense) (None, 256) 295168
activation_18 (Activation) (None, 256) 0
dropout_16 (Dropout) (None, 256) 0
dense_19 (Dense) (None, 256) 65792
activation_19 (Activation) (None, 256) 0
dropout_17 (Dropout) (None, 256) 0
dense_20 (Dense) (None, 10) 2570
activation_20 (Activation) (None, 10) 0
=================================================================
Total params: 406842 (1.55 MB)
Trainable params: 406842 (1.55 MB)
Non-trainable params: 0 (0.00 Byte)
_________________________________________________________________
55/55 [==============================] - 1s 20ms/step - loss: 3.5678 - accuracy: 0.1065
Pre-training accuracy: 10.6468%
num_epochs = 200
num_batch_size = 256
start = datetime.now()
history = model.fit(x_train_2D,
y_train_2D,
batch_size=num_batch_size,
epochs=num_epochs,
validation_data=(x_test_2D, y_test_2D),
verbose=1)
duration = datetime.now() - start
print("Training completed in time: ", duration)
Epoch 1/200 28/28 [==============================] - 20s 666ms/step - loss: 5.2682 - accuracy: 0.1210 - val_loss: 2.2766 - val_accuracy: 0.1923 Epoch 2/200 28/28 [==============================] - 16s 583ms/step - loss: 2.2443 - accuracy: 0.1626 - val_loss: 2.2073 - val_accuracy: 0.2210 Epoch 3/200 28/28 [==============================] - 15s 554ms/step - loss: 2.1717 - accuracy: 0.1980 - val_loss: 2.1329 - val_accuracy: 0.2553 Epoch 4/200 28/28 [==============================] - 16s 556ms/step - loss: 2.1056 - accuracy: 0.2206 - val_loss: 2.0706 - val_accuracy: 0.2822 Epoch 5/200 28/28 [==============================] - 15s 554ms/step - loss: 2.0502 - accuracy: 0.2487 - val_loss: 2.0159 - val_accuracy: 0.2833 Epoch 6/200 28/28 [==============================] - 15s 554ms/step - loss: 1.9452 - accuracy: 0.2805 - val_loss: 1.9698 - val_accuracy: 0.2753 Epoch 7/200 28/28 [==============================] - 16s 563ms/step - loss: 1.8753 - accuracy: 0.3026 - val_loss: 1.9002 - val_accuracy: 0.3171 Epoch 8/200 28/28 [==============================] - 15s 551ms/step - loss: 1.8043 - accuracy: 0.3323 - val_loss: 1.8043 - val_accuracy: 0.3578 Epoch 9/200 28/28 [==============================] - 16s 569ms/step - loss: 1.7327 - accuracy: 0.3522 - val_loss: 1.7456 - val_accuracy: 0.3892 Epoch 10/200 28/28 [==============================] - 16s 556ms/step - loss: 1.6744 - accuracy: 0.3888 - val_loss: 1.6930 - val_accuracy: 0.4184 Epoch 11/200 28/28 [==============================] - 18s 636ms/step - loss: 1.5930 - accuracy: 0.4180 - val_loss: 1.5688 - val_accuracy: 0.4900 Epoch 12/200 28/28 [==============================] - 16s 587ms/step - loss: 1.5385 - accuracy: 0.4339 - val_loss: 1.4900 - val_accuracy: 0.5117 Epoch 13/200 28/28 [==============================] - 18s 638ms/step - loss: 1.5057 - accuracy: 0.4598 - val_loss: 1.4484 - val_accuracy: 0.5026 Epoch 14/200 28/28 [==============================] - 16s 566ms/step - loss: 1.4332 - accuracy: 0.4760 - val_loss: 1.4313 - val_accuracy: 0.5072 Epoch 15/200 28/28 [==============================] - 17s 601ms/step - loss: 1.3902 - accuracy: 0.5032 - val_loss: 1.3649 - val_accuracy: 0.5386 Epoch 16/200 28/28 [==============================] - 22s 776ms/step - loss: 1.3485 - accuracy: 0.5178 - val_loss: 1.3413 - val_accuracy: 0.5495 Epoch 17/200 28/28 [==============================] - 25s 915ms/step - loss: 1.3098 - accuracy: 0.5361 - val_loss: 1.2577 - val_accuracy: 0.6027 Epoch 18/200 28/28 [==============================] - 29s 1s/step - loss: 1.2899 - accuracy: 0.5442 - val_loss: 1.2281 - val_accuracy: 0.6045 Epoch 19/200 28/28 [==============================] - 22s 777ms/step - loss: 1.2351 - accuracy: 0.5682 - val_loss: 1.2024 - val_accuracy: 0.6113 Epoch 20/200 28/28 [==============================] - 27s 954ms/step - loss: 1.2017 - accuracy: 0.5757 - val_loss: 1.1937 - val_accuracy: 0.5913 Epoch 21/200 28/28 [==============================] - 23s 832ms/step - loss: 1.1805 - accuracy: 0.5895 - val_loss: 1.1377 - val_accuracy: 0.6234 Epoch 22/200 28/28 [==============================] - 27s 975ms/step - loss: 1.1475 - accuracy: 0.6001 - val_loss: 1.1102 - val_accuracy: 0.6319 Epoch 23/200 28/28 [==============================] - 25s 891ms/step - loss: 1.1042 - accuracy: 0.6120 - val_loss: 1.0592 - val_accuracy: 0.6474 Epoch 24/200 28/28 [==============================] - 24s 861ms/step - loss: 1.0864 - accuracy: 0.6225 - val_loss: 1.0364 - val_accuracy: 0.6588 Epoch 25/200 28/28 [==============================] - 24s 840ms/step - loss: 1.0360 - accuracy: 0.6445 - val_loss: 1.0354 - val_accuracy: 0.6428 Epoch 26/200 28/28 [==============================] - 19s 664ms/step - loss: 1.0284 - accuracy: 0.6445 - val_loss: 1.0007 - val_accuracy: 0.6789 Epoch 27/200 28/28 [==============================] - 18s 632ms/step - loss: 1.0125 - accuracy: 0.6520 - val_loss: 0.9817 - val_accuracy: 0.6869 Epoch 28/200 28/28 [==============================] - 17s 609ms/step - loss: 0.9663 - accuracy: 0.6746 - val_loss: 0.9559 - val_accuracy: 0.6800 Epoch 29/200 28/28 [==============================] - 17s 613ms/step - loss: 0.9404 - accuracy: 0.6775 - val_loss: 0.9198 - val_accuracy: 0.6983 Epoch 30/200 28/28 [==============================] - 19s 667ms/step - loss: 0.9341 - accuracy: 0.6835 - val_loss: 0.9175 - val_accuracy: 0.6880 Epoch 31/200 28/28 [==============================] - 18s 661ms/step - loss: 0.9000 - accuracy: 0.6972 - val_loss: 0.8976 - val_accuracy: 0.6961 Epoch 32/200 28/28 [==============================] - 17s 608ms/step - loss: 0.8588 - accuracy: 0.7087 - val_loss: 0.8131 - val_accuracy: 0.7281 Epoch 33/200 28/28 [==============================] - 19s 675ms/step - loss: 0.8467 - accuracy: 0.7074 - val_loss: 0.8348 - val_accuracy: 0.7172 Epoch 34/200 28/28 [==============================] - 19s 668ms/step - loss: 0.7971 - accuracy: 0.7307 - val_loss: 0.8129 - val_accuracy: 0.7315 Epoch 35/200 28/28 [==============================] - 17s 611ms/step - loss: 0.8043 - accuracy: 0.7278 - val_loss: 0.7437 - val_accuracy: 0.7521 Epoch 36/200 28/28 [==============================] - 17s 618ms/step - loss: 0.7775 - accuracy: 0.7374 - val_loss: 0.7834 - val_accuracy: 0.7333 Epoch 37/200 28/28 [==============================] - 17s 610ms/step - loss: 0.7463 - accuracy: 0.7493 - val_loss: 0.7832 - val_accuracy: 0.7235 Epoch 38/200 28/28 [==============================] - 19s 688ms/step - loss: 0.7387 - accuracy: 0.7538 - val_loss: 0.7311 - val_accuracy: 0.7550 Epoch 39/200 28/28 [==============================] - 17s 615ms/step - loss: 0.7119 - accuracy: 0.7621 - val_loss: 0.6791 - val_accuracy: 0.7722 Epoch 40/200 28/28 [==============================] - 19s 690ms/step - loss: 0.6816 - accuracy: 0.7682 - val_loss: 0.6713 - val_accuracy: 0.7716 Epoch 41/200 28/28 [==============================] - 19s 664ms/step - loss: 0.6729 - accuracy: 0.7735 - val_loss: 0.6884 - val_accuracy: 0.7624 Epoch 42/200 28/28 [==============================] - 18s 652ms/step - loss: 0.6800 - accuracy: 0.7704 - val_loss: 0.6811 - val_accuracy: 0.7705 Epoch 43/200 28/28 [==============================] - 18s 637ms/step - loss: 0.6410 - accuracy: 0.7843 - val_loss: 0.6386 - val_accuracy: 0.7888 Epoch 44/200 28/28 [==============================] - 20s 731ms/step - loss: 0.6411 - accuracy: 0.7851 - val_loss: 0.6268 - val_accuracy: 0.7934 Epoch 45/200 28/28 [==============================] - 18s 656ms/step - loss: 0.6275 - accuracy: 0.7870 - val_loss: 0.6303 - val_accuracy: 0.7899 Epoch 46/200 28/28 [==============================] - 22s 799ms/step - loss: 0.6006 - accuracy: 0.8033 - val_loss: 0.5842 - val_accuracy: 0.8117 Epoch 47/200 28/28 [==============================] - 27s 947ms/step - loss: 0.5940 - accuracy: 0.8036 - val_loss: 0.5823 - val_accuracy: 0.8037 Epoch 48/200 28/28 [==============================] - 28s 1s/step - loss: 0.5896 - accuracy: 0.8037 - val_loss: 0.6105 - val_accuracy: 0.7871 Epoch 49/200 28/28 [==============================] - 27s 966ms/step - loss: 0.5938 - accuracy: 0.7993 - val_loss: 0.5975 - val_accuracy: 0.8042 Epoch 50/200 28/28 [==============================] - 29s 1s/step - loss: 0.5447 - accuracy: 0.8185 - val_loss: 0.6088 - val_accuracy: 0.7991 Epoch 51/200 28/28 [==============================] - 26s 940ms/step - loss: 0.5285 - accuracy: 0.8203 - val_loss: 0.5580 - val_accuracy: 0.8157 Epoch 52/200 28/28 [==============================] - 28s 995ms/step - loss: 0.5362 - accuracy: 0.8209 - val_loss: 0.5751 - val_accuracy: 0.8094 Epoch 53/200 28/28 [==============================] - 26s 939ms/step - loss: 0.5283 - accuracy: 0.8195 - val_loss: 0.5584 - val_accuracy: 0.8088 Epoch 54/200 28/28 [==============================] - 27s 981ms/step - loss: 0.5072 - accuracy: 0.8338 - val_loss: 0.5224 - val_accuracy: 0.8203 Epoch 55/200 28/28 [==============================] - 27s 974ms/step - loss: 0.5071 - accuracy: 0.8279 - val_loss: 0.5335 - val_accuracy: 0.8128 Epoch 56/200 28/28 [==============================] - 30s 1s/step - loss: 0.5021 - accuracy: 0.8309 - val_loss: 0.5385 - val_accuracy: 0.8157 Epoch 57/200 28/28 [==============================] - 27s 949ms/step - loss: 0.4787 - accuracy: 0.8381 - val_loss: 0.5100 - val_accuracy: 0.8226 Epoch 58/200 28/28 [==============================] - 31s 1s/step - loss: 0.4641 - accuracy: 0.8491 - val_loss: 0.5132 - val_accuracy: 0.8254 Epoch 59/200 28/28 [==============================] - 24s 848ms/step - loss: 0.4723 - accuracy: 0.8395 - val_loss: 0.5301 - val_accuracy: 0.8254 Epoch 60/200 28/28 [==============================] - 23s 835ms/step - loss: 0.4605 - accuracy: 0.8455 - val_loss: 0.5058 - val_accuracy: 0.8317 Epoch 61/200 28/28 [==============================] - 28s 1s/step - loss: 0.4571 - accuracy: 0.8462 - val_loss: 0.4722 - val_accuracy: 0.8443 Epoch 62/200 28/28 [==============================] - 24s 836ms/step - loss: 0.4267 - accuracy: 0.8553 - val_loss: 0.4744 - val_accuracy: 0.8454 Epoch 63/200 28/28 [==============================] - 23s 821ms/step - loss: 0.4366 - accuracy: 0.8545 - val_loss: 0.4502 - val_accuracy: 0.8454 Epoch 64/200 28/28 [==============================] - 25s 906ms/step - loss: 0.4339 - accuracy: 0.8527 - val_loss: 0.4484 - val_accuracy: 0.8535 Epoch 65/200 28/28 [==============================] - 21s 751ms/step - loss: 0.4047 - accuracy: 0.8603 - val_loss: 0.4542 - val_accuracy: 0.8449 Epoch 66/200 28/28 [==============================] - 20s 706ms/step - loss: 0.4231 - accuracy: 0.8588 - val_loss: 0.4940 - val_accuracy: 0.8426 Epoch 67/200 28/28 [==============================] - 21s 736ms/step - loss: 0.3845 - accuracy: 0.8700 - val_loss: 0.4596 - val_accuracy: 0.8603 Epoch 68/200 28/28 [==============================] - 26s 925ms/step - loss: 0.3859 - accuracy: 0.8693 - val_loss: 0.4406 - val_accuracy: 0.8495 Epoch 69/200 28/28 [==============================] - 20s 712ms/step - loss: 0.3840 - accuracy: 0.8710 - val_loss: 0.4347 - val_accuracy: 0.8523 Epoch 70/200 28/28 [==============================] - 17s 625ms/step - loss: 0.3750 - accuracy: 0.8742 - val_loss: 0.4471 - val_accuracy: 0.8569 Epoch 71/200 28/28 [==============================] - 26s 929ms/step - loss: 0.3624 - accuracy: 0.8816 - val_loss: 0.4504 - val_accuracy: 0.8443 Epoch 72/200 28/28 [==============================] - 24s 856ms/step - loss: 0.3818 - accuracy: 0.8730 - val_loss: 0.4672 - val_accuracy: 0.8426 Epoch 73/200 28/28 [==============================] - 27s 971ms/step - loss: 0.3587 - accuracy: 0.8805 - val_loss: 0.4309 - val_accuracy: 0.8569 Epoch 74/200 28/28 [==============================] - 27s 945ms/step - loss: 0.3512 - accuracy: 0.8842 - val_loss: 0.4342 - val_accuracy: 0.8506 Epoch 75/200 28/28 [==============================] - 41s 1s/step - loss: 0.3309 - accuracy: 0.8882 - val_loss: 0.3838 - val_accuracy: 0.8758 Epoch 76/200 28/28 [==============================] - 22s 757ms/step - loss: 0.3419 - accuracy: 0.8820 - val_loss: 0.4485 - val_accuracy: 0.8552 Epoch 77/200 28/28 [==============================] - 19s 686ms/step - loss: 0.3531 - accuracy: 0.8889 - val_loss: 0.4286 - val_accuracy: 0.8580 Epoch 78/200 28/28 [==============================] - 25s 880ms/step - loss: 0.3321 - accuracy: 0.8905 - val_loss: 0.4121 - val_accuracy: 0.8603 Epoch 79/200 28/28 [==============================] - 24s 848ms/step - loss: 0.3287 - accuracy: 0.8899 - val_loss: 0.3846 - val_accuracy: 0.8735 Epoch 80/200 28/28 [==============================] - 21s 751ms/step - loss: 0.3273 - accuracy: 0.8913 - val_loss: 0.4067 - val_accuracy: 0.8655 Epoch 81/200 28/28 [==============================] - 20s 703ms/step - loss: 0.3164 - accuracy: 0.8925 - val_loss: 0.3721 - val_accuracy: 0.8706 Epoch 82/200 28/28 [==============================] - 17s 589ms/step - loss: 0.3201 - accuracy: 0.8923 - val_loss: 0.4152 - val_accuracy: 0.8718 Epoch 83/200 28/28 [==============================] - 16s 575ms/step - loss: 0.3137 - accuracy: 0.8961 - val_loss: 0.3870 - val_accuracy: 0.8746 Epoch 84/200 28/28 [==============================] - 16s 566ms/step - loss: 0.3133 - accuracy: 0.8926 - val_loss: 0.3720 - val_accuracy: 0.8764 Epoch 85/200 28/28 [==============================] - 16s 569ms/step - loss: 0.3064 - accuracy: 0.8935 - val_loss: 0.4092 - val_accuracy: 0.8626 Epoch 86/200 28/28 [==============================] - 16s 563ms/step - loss: 0.2956 - accuracy: 0.8985 - val_loss: 0.3714 - val_accuracy: 0.8769 Epoch 87/200 28/28 [==============================] - 19s 673ms/step - loss: 0.3015 - accuracy: 0.8999 - val_loss: 0.3854 - val_accuracy: 0.8746 Epoch 88/200 28/28 [==============================] - 18s 612ms/step - loss: 0.2939 - accuracy: 0.9006 - val_loss: 0.3767 - val_accuracy: 0.8695 Epoch 89/200 28/28 [==============================] - 16s 562ms/step - loss: 0.2716 - accuracy: 0.9097 - val_loss: 0.3615 - val_accuracy: 0.8895 Epoch 90/200 28/28 [==============================] - 15s 551ms/step - loss: 0.2778 - accuracy: 0.9068 - val_loss: 0.3446 - val_accuracy: 0.8941 Epoch 91/200 28/28 [==============================] - 16s 558ms/step - loss: 0.2702 - accuracy: 0.9097 - val_loss: 0.3584 - val_accuracy: 0.8849 Epoch 92/200 28/28 [==============================] - 16s 554ms/step - loss: 0.2826 - accuracy: 0.9058 - val_loss: 0.3243 - val_accuracy: 0.8924 Epoch 93/200 28/28 [==============================] - 16s 559ms/step - loss: 0.2607 - accuracy: 0.9124 - val_loss: 0.3638 - val_accuracy: 0.8809 Epoch 94/200 28/28 [==============================] - 15s 548ms/step - loss: 0.2724 - accuracy: 0.9110 - val_loss: 0.3820 - val_accuracy: 0.8718 Epoch 95/200 28/28 [==============================] - 15s 551ms/step - loss: 0.2785 - accuracy: 0.9042 - val_loss: 0.3920 - val_accuracy: 0.8769 Epoch 96/200 28/28 [==============================] - 15s 548ms/step - loss: 0.2485 - accuracy: 0.9157 - val_loss: 0.3559 - val_accuracy: 0.8838 Epoch 97/200 28/28 [==============================] - 16s 557ms/step - loss: 0.2557 - accuracy: 0.9137 - val_loss: 0.3480 - val_accuracy: 0.8798 Epoch 98/200 28/28 [==============================] - 15s 548ms/step - loss: 0.2786 - accuracy: 0.9059 - val_loss: 0.3629 - val_accuracy: 0.8832 Epoch 99/200 28/28 [==============================] - 16s 555ms/step - loss: 0.2371 - accuracy: 0.9197 - val_loss: 0.3368 - val_accuracy: 0.8867 Epoch 100/200 28/28 [==============================] - 15s 550ms/step - loss: 0.2303 - accuracy: 0.9184 - val_loss: 0.3363 - val_accuracy: 0.8895 Epoch 101/200 28/28 [==============================] - 16s 569ms/step - loss: 0.2481 - accuracy: 0.9191 - val_loss: 0.3710 - val_accuracy: 0.8844 Epoch 102/200 28/28 [==============================] - 16s 561ms/step - loss: 0.2423 - accuracy: 0.9181 - val_loss: 0.3207 - val_accuracy: 0.9044 Epoch 103/200 28/28 [==============================] - 15s 546ms/step - loss: 0.2318 - accuracy: 0.9187 - val_loss: 0.3261 - val_accuracy: 0.9015 Epoch 104/200 28/28 [==============================] - 16s 555ms/step - loss: 0.2291 - accuracy: 0.9246 - val_loss: 0.3103 - val_accuracy: 0.8998 Epoch 105/200 28/28 [==============================] - 16s 561ms/step - loss: 0.2186 - accuracy: 0.9225 - val_loss: 0.3275 - val_accuracy: 0.8998 Epoch 106/200 28/28 [==============================] - 15s 550ms/step - loss: 0.2386 - accuracy: 0.9233 - val_loss: 0.3194 - val_accuracy: 0.8964 Epoch 107/200 28/28 [==============================] - 16s 562ms/step - loss: 0.2321 - accuracy: 0.9203 - val_loss: 0.3172 - val_accuracy: 0.8981 Epoch 108/200 28/28 [==============================] - 17s 602ms/step - loss: 0.2097 - accuracy: 0.9303 - val_loss: 0.3295 - val_accuracy: 0.8947 Epoch 109/200 28/28 [==============================] - 16s 575ms/step - loss: 0.2015 - accuracy: 0.9306 - val_loss: 0.3363 - val_accuracy: 0.8958 Epoch 110/200 28/28 [==============================] - 16s 564ms/step - loss: 0.2098 - accuracy: 0.9307 - val_loss: 0.3087 - val_accuracy: 0.9033 Epoch 111/200 28/28 [==============================] - 15s 552ms/step - loss: 0.2094 - accuracy: 0.9301 - val_loss: 0.3381 - val_accuracy: 0.8952 Epoch 112/200 28/28 [==============================] - 16s 561ms/step - loss: 0.2006 - accuracy: 0.9336 - val_loss: 0.3310 - val_accuracy: 0.8975 Epoch 113/200 28/28 [==============================] - 15s 545ms/step - loss: 0.2032 - accuracy: 0.9311 - val_loss: 0.3243 - val_accuracy: 0.9004 Epoch 114/200 28/28 [==============================] - 15s 546ms/step - loss: 0.2181 - accuracy: 0.9273 - val_loss: 0.3581 - val_accuracy: 0.8884 Epoch 115/200 28/28 [==============================] - 15s 551ms/step - loss: 0.2064 - accuracy: 0.9330 - val_loss: 0.3236 - val_accuracy: 0.8998 Epoch 116/200 28/28 [==============================] - 15s 544ms/step - loss: 0.1998 - accuracy: 0.9331 - val_loss: 0.3351 - val_accuracy: 0.8964 Epoch 117/200 28/28 [==============================] - 15s 542ms/step - loss: 0.1864 - accuracy: 0.9353 - val_loss: 0.3185 - val_accuracy: 0.9021 Epoch 118/200 28/28 [==============================] - 15s 546ms/step - loss: 0.2026 - accuracy: 0.9331 - val_loss: 0.3181 - val_accuracy: 0.8998 Epoch 119/200 28/28 [==============================] - 15s 547ms/step - loss: 0.2089 - accuracy: 0.9287 - val_loss: 0.3094 - val_accuracy: 0.9061 Epoch 120/200 28/28 [==============================] - 15s 553ms/step - loss: 0.1820 - accuracy: 0.9407 - val_loss: 0.3048 - val_accuracy: 0.9078 Epoch 121/200 28/28 [==============================] - 16s 556ms/step - loss: 0.1872 - accuracy: 0.9400 - val_loss: 0.3170 - val_accuracy: 0.9021 Epoch 122/200 28/28 [==============================] - 15s 554ms/step - loss: 0.1842 - accuracy: 0.9377 - val_loss: 0.3445 - val_accuracy: 0.8884 Epoch 123/200 28/28 [==============================] - 16s 563ms/step - loss: 0.1849 - accuracy: 0.9387 - val_loss: 0.3267 - val_accuracy: 0.8970 Epoch 124/200 28/28 [==============================] - 16s 554ms/step - loss: 0.1970 - accuracy: 0.9320 - val_loss: 0.3592 - val_accuracy: 0.8918 Epoch 125/200 28/28 [==============================] - 16s 562ms/step - loss: 0.2056 - accuracy: 0.9327 - val_loss: 0.3170 - val_accuracy: 0.9038 Epoch 126/200 28/28 [==============================] - 16s 555ms/step - loss: 0.1926 - accuracy: 0.9311 - val_loss: 0.3138 - val_accuracy: 0.9021 Epoch 127/200 28/28 [==============================] - 470s 17s/step - loss: 0.1908 - accuracy: 0.9393 - val_loss: 0.3150 - val_accuracy: 0.8998 Epoch 128/200 28/28 [==============================] - 22s 774ms/step - loss: 0.1834 - accuracy: 0.9416 - val_loss: 0.2986 - val_accuracy: 0.9027 Epoch 129/200 28/28 [==============================] - 21s 772ms/step - loss: 0.1669 - accuracy: 0.9426 - val_loss: 0.3071 - val_accuracy: 0.9124 Epoch 130/200 28/28 [==============================] - 18s 657ms/step - loss: 0.1743 - accuracy: 0.9416 - val_loss: 0.2900 - val_accuracy: 0.9141 Epoch 131/200 28/28 [==============================] - 17s 616ms/step - loss: 0.1735 - accuracy: 0.9440 - val_loss: 0.2973 - val_accuracy: 0.9067 Epoch 132/200 28/28 [==============================] - 16s 573ms/step - loss: 0.1717 - accuracy: 0.9435 - val_loss: 0.2948 - val_accuracy: 0.9096 Epoch 133/200 28/28 [==============================] - 16s 567ms/step - loss: 0.1682 - accuracy: 0.9442 - val_loss: 0.3092 - val_accuracy: 0.9056 Epoch 134/200 28/28 [==============================] - 16s 572ms/step - loss: 0.1687 - accuracy: 0.9443 - val_loss: 0.2727 - val_accuracy: 0.9141 Epoch 135/200 28/28 [==============================] - 16s 581ms/step - loss: 0.1711 - accuracy: 0.9403 - val_loss: 0.3028 - val_accuracy: 0.9090 Epoch 136/200 28/28 [==============================] - 16s 576ms/step - loss: 0.1670 - accuracy: 0.9416 - val_loss: 0.3131 - val_accuracy: 0.9038 Epoch 137/200 28/28 [==============================] - 16s 572ms/step - loss: 0.1629 - accuracy: 0.9443 - val_loss: 0.3189 - val_accuracy: 0.9124 Epoch 138/200 28/28 [==============================] - 16s 589ms/step - loss: 0.1581 - accuracy: 0.9475 - val_loss: 0.2830 - val_accuracy: 0.9073 Epoch 139/200 28/28 [==============================] - 17s 595ms/step - loss: 0.1617 - accuracy: 0.9476 - val_loss: 0.3095 - val_accuracy: 0.9090 Epoch 140/200 28/28 [==============================] - 17s 610ms/step - loss: 0.1561 - accuracy: 0.9443 - val_loss: 0.3491 - val_accuracy: 0.8964 Epoch 141/200 28/28 [==============================] - 16s 588ms/step - loss: 0.1580 - accuracy: 0.9473 - val_loss: 0.3564 - val_accuracy: 0.8901 Epoch 142/200 28/28 [==============================] - 16s 567ms/step - loss: 0.1664 - accuracy: 0.9446 - val_loss: 0.3271 - val_accuracy: 0.8912 Epoch 143/200 28/28 [==============================] - 15s 542ms/step - loss: 0.1735 - accuracy: 0.9433 - val_loss: 0.2726 - val_accuracy: 0.9107 Epoch 144/200 28/28 [==============================] - 16s 561ms/step - loss: 0.1559 - accuracy: 0.9485 - val_loss: 0.3020 - val_accuracy: 0.9050 Epoch 145/200 28/28 [==============================] - 15s 548ms/step - loss: 0.1439 - accuracy: 0.9508 - val_loss: 0.3088 - val_accuracy: 0.9073 Epoch 146/200 28/28 [==============================] - 15s 555ms/step - loss: 0.1592 - accuracy: 0.9459 - val_loss: 0.3260 - val_accuracy: 0.8952 Epoch 147/200 28/28 [==============================] - 15s 540ms/step - loss: 0.1441 - accuracy: 0.9485 - val_loss: 0.2810 - val_accuracy: 0.9130 Epoch 148/200 28/28 [==============================] - 15s 541ms/step - loss: 0.1518 - accuracy: 0.9515 - val_loss: 0.2821 - val_accuracy: 0.9096 Epoch 149/200 28/28 [==============================] - 15s 549ms/step - loss: 0.1384 - accuracy: 0.9550 - val_loss: 0.3193 - val_accuracy: 0.9038 Epoch 150/200 28/28 [==============================] - 15s 547ms/step - loss: 0.1561 - accuracy: 0.9477 - val_loss: 0.3277 - val_accuracy: 0.8970 Epoch 151/200 28/28 [==============================] - 16s 572ms/step - loss: 0.1538 - accuracy: 0.9495 - val_loss: 0.3405 - val_accuracy: 0.9004 Epoch 152/200 28/28 [==============================] - 16s 579ms/step - loss: 0.1535 - accuracy: 0.9473 - val_loss: 0.2993 - val_accuracy: 0.9073 Epoch 153/200 28/28 [==============================] - 15s 550ms/step - loss: 0.1382 - accuracy: 0.9543 - val_loss: 0.2936 - val_accuracy: 0.9124 Epoch 154/200 28/28 [==============================] - 15s 546ms/step - loss: 0.1380 - accuracy: 0.9509 - val_loss: 0.2868 - val_accuracy: 0.9181 Epoch 155/200 28/28 [==============================] - 15s 544ms/step - loss: 0.1365 - accuracy: 0.9528 - val_loss: 0.3167 - val_accuracy: 0.9130 Epoch 156/200 28/28 [==============================] - 16s 555ms/step - loss: 0.1327 - accuracy: 0.9560 - val_loss: 0.2941 - val_accuracy: 0.9193 Epoch 157/200 28/28 [==============================] - 15s 554ms/step - loss: 0.1494 - accuracy: 0.9523 - val_loss: 0.3071 - val_accuracy: 0.9090 Epoch 158/200 28/28 [==============================] - 16s 579ms/step - loss: 0.1617 - accuracy: 0.9502 - val_loss: 0.3005 - val_accuracy: 0.9141 Epoch 159/200 28/28 [==============================] - 15s 541ms/step - loss: 0.1334 - accuracy: 0.9546 - val_loss: 0.3275 - val_accuracy: 0.9096 Epoch 160/200 28/28 [==============================] - 15s 544ms/step - loss: 0.1502 - accuracy: 0.9515 - val_loss: 0.2923 - val_accuracy: 0.9124 Epoch 161/200 28/28 [==============================] - 15s 543ms/step - loss: 0.1428 - accuracy: 0.9546 - val_loss: 0.3140 - val_accuracy: 0.9056 Epoch 162/200 28/28 [==============================] - 15s 544ms/step - loss: 0.1250 - accuracy: 0.9602 - val_loss: 0.2890 - val_accuracy: 0.9187 Epoch 163/200 28/28 [==============================] - 15s 541ms/step - loss: 0.1278 - accuracy: 0.9545 - val_loss: 0.2779 - val_accuracy: 0.9130 Epoch 164/200 28/28 [==============================] - 15s 542ms/step - loss: 0.1327 - accuracy: 0.9549 - val_loss: 0.3050 - val_accuracy: 0.9113 Epoch 165/200 28/28 [==============================] - 16s 559ms/step - loss: 0.1243 - accuracy: 0.9586 - val_loss: 0.2837 - val_accuracy: 0.9159 Epoch 166/200 28/28 [==============================] - 15s 546ms/step - loss: 0.1346 - accuracy: 0.9552 - val_loss: 0.2598 - val_accuracy: 0.9244 Epoch 167/200 28/28 [==============================] - 15s 545ms/step - loss: 0.1333 - accuracy: 0.9578 - val_loss: 0.3137 - val_accuracy: 0.9073 Epoch 168/200 28/28 [==============================] - 15s 544ms/step - loss: 0.1164 - accuracy: 0.9612 - val_loss: 0.2762 - val_accuracy: 0.9153 Epoch 169/200 28/28 [==============================] - 15s 546ms/step - loss: 0.1297 - accuracy: 0.9575 - val_loss: 0.2606 - val_accuracy: 0.9222 Epoch 170/200 28/28 [==============================] - 667s 25s/step - loss: 0.1293 - accuracy: 0.9540 - val_loss: 0.2969 - val_accuracy: 0.9101 Epoch 171/200 28/28 [==============================] - 22s 797ms/step - loss: 0.1324 - accuracy: 0.9553 - val_loss: 0.2921 - val_accuracy: 0.9187 Epoch 172/200 28/28 [==============================] - 16s 563ms/step - loss: 0.1273 - accuracy: 0.9565 - val_loss: 0.2622 - val_accuracy: 0.9210 Epoch 173/200 28/28 [==============================] - 16s 556ms/step - loss: 0.1237 - accuracy: 0.9606 - val_loss: 0.2671 - val_accuracy: 0.9262 Epoch 174/200 28/28 [==============================] - 16s 561ms/step - loss: 0.1173 - accuracy: 0.9603 - val_loss: 0.2959 - val_accuracy: 0.9136 Epoch 175/200 28/28 [==============================] - 16s 554ms/step - loss: 0.1170 - accuracy: 0.9593 - val_loss: 0.2643 - val_accuracy: 0.9216 Epoch 176/200 28/28 [==============================] - 15s 546ms/step - loss: 0.1244 - accuracy: 0.9601 - val_loss: 0.3103 - val_accuracy: 0.9130 Epoch 177/200 28/28 [==============================] - 15s 551ms/step - loss: 0.1288 - accuracy: 0.9593 - val_loss: 0.3006 - val_accuracy: 0.9164 Epoch 178/200 28/28 [==============================] - 15s 552ms/step - loss: 0.1235 - accuracy: 0.9573 - val_loss: 0.2679 - val_accuracy: 0.9244 Epoch 179/200 28/28 [==============================] - 15s 546ms/step - loss: 0.1153 - accuracy: 0.9601 - val_loss: 0.2866 - val_accuracy: 0.9216 Epoch 180/200 28/28 [==============================] - 15s 545ms/step - loss: 0.1101 - accuracy: 0.9618 - val_loss: 0.2765 - val_accuracy: 0.9256 Epoch 181/200 28/28 [==============================] - 15s 547ms/step - loss: 0.1261 - accuracy: 0.9593 - val_loss: 0.2697 - val_accuracy: 0.9244 Epoch 182/200 28/28 [==============================] - 15s 545ms/step - loss: 0.1193 - accuracy: 0.9609 - val_loss: 0.3223 - val_accuracy: 0.9113 Epoch 183/200 28/28 [==============================] - 16s 557ms/step - loss: 0.1112 - accuracy: 0.9644 - val_loss: 0.3197 - val_accuracy: 0.9090 Epoch 184/200 28/28 [==============================] - 15s 545ms/step - loss: 0.1120 - accuracy: 0.9652 - val_loss: 0.3549 - val_accuracy: 0.9078 Epoch 185/200 28/28 [==============================] - 15s 543ms/step - loss: 0.1142 - accuracy: 0.9644 - val_loss: 0.2738 - val_accuracy: 0.9227 Epoch 186/200 28/28 [==============================] - 15s 548ms/step - loss: 0.1137 - accuracy: 0.9602 - val_loss: 0.2755 - val_accuracy: 0.9279 Epoch 187/200 28/28 [==============================] - 15s 541ms/step - loss: 0.1406 - accuracy: 0.9569 - val_loss: 0.2814 - val_accuracy: 0.9222 Epoch 188/200 28/28 [==============================] - 15s 542ms/step - loss: 0.1150 - accuracy: 0.9613 - val_loss: 0.2985 - val_accuracy: 0.9176 Epoch 189/200 28/28 [==============================] - 15s 546ms/step - loss: 0.1096 - accuracy: 0.9648 - val_loss: 0.2774 - val_accuracy: 0.9153 Epoch 190/200 28/28 [==============================] - 15s 540ms/step - loss: 0.1092 - accuracy: 0.9656 - val_loss: 0.2635 - val_accuracy: 0.9267 Epoch 191/200 28/28 [==============================] - 15s 542ms/step - loss: 0.1057 - accuracy: 0.9646 - val_loss: 0.3240 - val_accuracy: 0.9096 Epoch 192/200 28/28 [==============================] - 15s 548ms/step - loss: 0.1093 - accuracy: 0.9634 - val_loss: 0.3133 - val_accuracy: 0.9124 Epoch 193/200 28/28 [==============================] - 15s 552ms/step - loss: 0.1140 - accuracy: 0.9639 - val_loss: 0.2903 - val_accuracy: 0.9193 Epoch 194/200 28/28 [==============================] - 15s 553ms/step - loss: 0.1096 - accuracy: 0.9646 - val_loss: 0.2983 - val_accuracy: 0.9124 Epoch 195/200 28/28 [==============================] - 15s 548ms/step - loss: 0.1011 - accuracy: 0.9674 - val_loss: 0.2762 - val_accuracy: 0.9239 Epoch 196/200 28/28 [==============================] - 16s 576ms/step - loss: 0.1009 - accuracy: 0.9674 - val_loss: 0.2894 - val_accuracy: 0.9244 Epoch 197/200 28/28 [==============================] - 16s 563ms/step - loss: 0.1073 - accuracy: 0.9642 - val_loss: 0.2834 - val_accuracy: 0.9170 Epoch 198/200 28/28 [==============================] - 15s 547ms/step - loss: 0.1050 - accuracy: 0.9656 - val_loss: 0.3541 - val_accuracy: 0.9021 Epoch 199/200 28/28 [==============================] - 15s 548ms/step - loss: 0.1066 - accuracy: 0.9652 - val_loss: 0.2713 - val_accuracy: 0.9239 Epoch 200/200 28/28 [==============================] - 15s 544ms/step - loss: 0.0989 - accuracy: 0.9684 - val_loss: 0.2938 - val_accuracy: 0.9170 Training completed in time: 1:19:05.880745
# Evaluating the model on the training and testing set
cnn_acc_train = model.evaluate(x_train_2D, y_train_2D, verbose=0)
print("Training Accuracy: ", cnn_acc_train[1])
cnn_acc_test = model.evaluate(x_test_2D, y_test_2D, verbose=0)
print("Testing Accuracy: ", cnn_acc_test[1])
Training Accuracy: 0.9958482384681702 Testing Accuracy: 0.9170005917549133
plot_accuracy(history,num_epochs,num_batch_size)
plot_losses(history,num_epochs)
model_json = model.to_json()
with open("model.json", "w") as json_file:
json_file.write(model_json)
# serialize weights to HDF5
model.save_weights("model.h5")
# load json and create model
json_file = open('model.json', 'r')
loaded_model_json = json_file.read()
json_file.close()
loaded_model = model_from_json(loaded_model_json)
# load weights into new model
loaded_model.load_weights("model.h5")
loaded_model.compile(loss='categorical_crossentropy', metrics=['accuracy'], optimizer='adam')
score = loaded_model.evaluate(x_test_2D, y_test_2D, verbose=0)
print("%s: %.2f%%" % (loaded_model.metrics_names[1], score[1]*100))
accuracy: 91.70%
y_pred_2D = model.predict(x_test_2D)
55/55 [==============================] - 3s 24ms/step
def extract_features(file_name, max_pad=174, n=40):
try:
audio, sample_rate = librosa.load(file_name, res_type='kaiser_fast')
mfccs = librosa.feature.mfcc(y=audio, sr=sample_rate, n_mfcc=n)
# padding
pad_width = max_pad - mfccs.shape[1]
mfccs_pad = np.pad(mfccs, pad_width=((0, 0), (0, pad_width)), mode='constant')
mfccsscaled = np.mean(mfccs.T, axis=0)
except Exception as e:
print("Error encountered while parsing file:", file_name)
return None
return mfccs, mfccs_pad, mfccsscaled
class_names= ['air_conditioner', 'car_horn','car_horn','children_playing','dog_bark','drilling','engine_idling','gun_shot','jackhammer','siren','street_music', ]
label_encoder_2D = LabelEncoder()
label_encoder_2D.fit(class_names)
def predict(file_name, max_pad=174, n=40):
pred_features, pred_features_pad, pred_features_scaled = extract_features(file_name, max_pad, n)
pred_feature = pred_features_pad.reshape(1, num_rows, num_columns, num_channels)
pred_v = model.predict(pred_feature)
preditcted_classes=np.argmax(pred_v, axis=1)
pred_classes = label_encoder_2D.inverse_transform(preditcted_classes)
print("The predicted class is:", pred_classes[0], '\n')
# predicted_proba_vector = model.predict_proba(pred_feature)
# predicted_proba = predicted_proba_vector[0]
# for i in range(len(predicted_proba)):
# category = le.inverse_transform(np.array([i]))
# print(category[0], "\t\t : ", format(predicted_proba[i], '.32f'))
predicted_proba = pred_v[0] # Get predicted probabilities directly from the predict method
for i, prob in enumerate(predicted_proba):
category = label_encoder_2D.inverse_transform(np.array([i]))
print(category[0], "\t\t : ", format(prob, '.32f'))
predict('E:/Projects/Audio Classification/Audio_Classification_4_using_models_(Urbansound8 dataset)/Sample_sound/jackhammer.wav', max_pad=174, n=40)
ipd.Audio('E:/Projects/Audio Classification/Audio_Classification_4_using_models_(Urbansound8 dataset)/Sample_sound/jackhammer.wav')
1/1 [==============================] - 0s 37ms/step The predicted class is: jackhammer air_conditioner : 0.00000085888774492559605278074741 car_horn : 0.00007206947339000180363655090332 children_playing : 0.00000000190029836311111921531847 dog_bark : 0.00000049399449153497698716819286 drilling : 0.00069468154106289148330688476562 engine_idling : 0.00000084055091065238229930400848 gun_shot : 0.00000000169058578247671675853780 jackhammer : 0.99922776222229003906250000000000 siren : 0.00000259413604908331762999296188 street_music : 0.00000073191415594919817522168159
predict('E:/Projects/Audio Classification/Audio_Classification_4_using_models_(Urbansound8 dataset)/Sample_sound/drilling.wav', max_pad=174, n=40)
ipd.Audio('E:/Projects/Audio Classification/Audio_Classification_4_using_models_(Urbansound8 dataset)/Sample_sound/drilling.wav')
1/1 [==============================] - 0s 39ms/step The predicted class is: drilling air_conditioner : 0.00000000000000335794904049121525 car_horn : 0.00000000000017995401989292364453 children_playing : 0.00000000000000001570934270065795 dog_bark : 0.00000000000011252517608017001405 drilling : 1.00000000000000000000000000000000 engine_idling : 0.00000000000000000362152121243934 gun_shot : 0.00000000000000195978796511703680 jackhammer : 0.00000000005411809542166245989847 siren : 0.00000000000005004095485964914214 street_music : 0.00000000000000000051503403054000
predict('E:/Projects/Audio Classification/Audio_Classification_4_using_models_(Urbansound8 dataset)/Audio_dataset/fold4/22883-7-71-0.wav', max_pad=174, n=40)
1/1 [==============================] - 0s 33ms/step The predicted class is: jackhammer air_conditioner : 0.00000048977466349242604337632656 car_horn : 0.00000689605349180055782198905945 children_playing : 0.00000000001413362309132804028877 dog_bark : 0.00000000074528205828983118408360 drilling : 0.00000039481821545450657140463591 engine_idling : 0.00000064945828626150614582002163 gun_shot : 0.00000000011368068292272326402781 jackhammer : 0.99999165534973144531250000000000 siren : 0.00000001686096595676644938066602 street_music : 0.00000000114699827413744515069993
## checking whether the path exists or not
os.path.exists('E:/Projects/Audio Classification/Audio_Classification_4_using_models_(Urbansound8 dataset)/electric-drill-01.wav')
True
# Convert one-hot encoded predictions back to class indices
class_names = meta_data.groupby(['class', 'classID'], as_index = False).sum()['class']
y_pred_indices = np.argmax(y_pred_2D, axis=1)
# Convert one-hot encoded true labels back to class indices
y_true_indices = np.argmax(y_test_2D, axis=1)
# Call the plot_confusion_matrix function with class indices
plot_confusion_matrix(y_true_indices, y_pred_indices, classes=class_names, cmap=plt.cm.Blues)
<Axes: title={'center': 'Confusion matrix, without normalization'}, xlabel='Predicted label', ylabel='True label'>
num_classes = len(np.unique(y_pred_indices))
print(num_classes)
10
| Model Name | Training Accuracy(%) | Testing Accuracy |
|---|---|---|
| Random Forest Classification | 57.12 | 53.34 |
| Support Vector Machine(SVM) | 99.87 | 89.81 |
| Multi-Layer Perceptron | 82.10 | 76.13 |
| Convolutional Neural Network | 99.58 | 91.70 |